Molecular recognition of β-O-GlcNAc glycopeptides by a lectin-like receptor: Binding modulation by the underlying ser or thr amino acids

Article abstract of DOI:10.1002/cbic.201000526

The binding properties of different carbohydrates and glycopeptides containing the β-O-2-deoxy-2-(N-acetyl)-D-glucosaminyl (β-O-GlcNAc) to a synthetically prepared lectin-like receptor have been analyzed. The study combines the use of NMR spectroscopy experiments with extensive MD simulations in explicit water. Notably, the presence of a key hydrogen bond between the receptor and the OMe group of the β-O-GlcNAc-OMe derivative appears to be responsible for the high selectivity observed for this compound. In addition, to study the effect on the binding of the underlying amino acid, we have prepared different model glycopeptides, which include the non-natural α-methylserine and α-methylthreonine as underlying amino acids. Interestingly, the presence of a methyl group decreases the affinity constant, especially in those cases in which a β-methyl group is present. As a result, the serine-containing glycopeptide exhibited the highest affinity constant of the glycopeptides, and the threonine derivative showed the lowest one. This low selectivity could have its origin in the difficulty to form both specific hydrogen bonds and hydrophobic (CH-π) contacts.The binding properties of different carbohydrates and glycopeptides with β-O-GlcNAc to a synthetically prepared lectin-like receptor have been analyzed. A serine-containing glycopeptide exhibited the highest affinity constant of the glycopeptides, and a threonine derivative showed the lowest one. This low selectivity could have its origin in the difficulty to form both specific hydrogen bonds and hydrophobic (CH-π) contacts.

Full text of DOI:10.1002/cbic.201000526

DOI: 10.1002/cbic.201000526
Molecular Recognition of b-O-GlcNAc Glycopeptides by a
Lectin-Like Receptor: Binding Modulation by the
Underlying Ser or Thr Amino Acids
Francisco Corzana,[a] Alberto Fernꢀndez-Tejada,[a] Jesffls H. Busto,[a] Gururaj Joshi,[b]
Anthony P. Davis,*[b] Jesffls JimØnez-Barbero,[c] Alberto Avenoza,[a] and Jesffls M. Peregrina*[a]
The binding properties of different carbohydrates and glyco-
peptides containing the b-O-2-deoxy-2-(N-acetyl)-d-glucosa-
minyl (b-O-GlcNAc) to a synthetically prepared lectin-like re-
ceptor have been analyzed. The study combines the use of
NMR spectroscopy experiments with extensive MD simulations
in explicit water. Notably, the presence of a key hydrogen
bond between the receptor and the OMe group of the b-O-
GlcNAc-OMe derivative appears to be responsible for the high
selectivity observed for this compound. In addition, to study
the effect on the binding of the underlying amino acid, we
have prepared different model glycopeptides, which include
the non-natural a-methylserine and a-methylthreonine as
underlying amino acids. Interestingly, the presence of a methyl
group decreases the affinity constant, especially in those cases
in which a b-methyl group is present. As a result, the serine-
containing glycopeptide exhibited the highest affinity constant
of the glycopeptides, and the threonine derivative showed the
lowest one. This low selectivity could have its origin in the dif-
ficulty to form both specific hydrogen bonds and hydrophobic
(CH-p) contacts.
Introduction
b-O-2-Deoxy-2-(N-acetyl)-d-glucosaminyl (b-O-GlcNAc) moiety
is a common motif in biological chemistry[1] (Scheme 1A). It
plays a unique role in protein regulation through linkage to
the hydroxy group of serine or threonine. This post-translatio-
nal[1c] modification is highly dynamic[2] and draws comparisons
with protein phosphorylation as a biological control mecha-
nism. It has been implicated in gene transcription, nuclear traf-
ficking, protein translation,[3] signal transduction,[1a] the regula-
tion of protein–protein interactions,[1] and the sensing of nutri-
tional levels within the cell.[4] Furthermore, there is clear evi-
dence that the aberrant O-GlcNAc modification of proteins is
correlated with diabetes, tumorgenesis, and even with Alz-
heimer’s disease.[1,5]
explore the role of aglycon structure in the molecular recogni-
tion of b-O-GlcNAc.
Results and Discussion
To gain insights into the receptor–ligand interactions that play
a pivotal role in the strong selectivity observed for compound
1, we decided to run 80 ns MD simulations with time-averaged
restraints (MD-tar)[7] on the receptor–1 complex by using all
the experimental distances[6] as restraints and by following our
previously reported protocol.[8] Figure 1A shows 15 snapshots
taken from these MD-tar simulations. In this particular case,
the calculations were carried out without the externally direct-
ed, water-solubilizing side chains. The obtained structure was
fairly similar to the NOE-based conformation previously report-
ed. Seven conventional intramolecular hydrogen bonds were
detected throughout the simulations. The oxygen of the NHAc
is involved in a hydrogen bond with NH5, and its NH group in-
Bearing in mind the singular features of b-O-GlcNAc, it is cru-
cial to know in detail the interaction of this carbohydrate
moiety with its biological targets. On this basis, some of us[6]
have recently reported on the synthesis of a simple lectin-like
receptor (Scheme 1B) that is particularly effective for this sub-
strate. In fact, the receptor binds 1 (a model for b-O-GlcNAc)
with an affinity constant (Ka) of 630mꢀ1, very similar to that
shown by the lectin wheat germ agglutinin. The receptor is
also highly selective for the b-anomer. Indeed, binding to relat-
ed carbohydrates is significantly weaker (e.g., Ka for 2 and a-
anomer 3 are 156 and 24mꢀ1, respectively) whereas affinities
[a] Dr. F. Corzana, Dr. A. Fernꢀndez-Tejada, Dr. J. H. Busto, Prof. A. Avenoza,
Prof. J. M. Peregrina
Departamento de Quꢁmica. Universidad de La Rioja, UA-CSIC
26006 LogroÇo (Spain)
Fax: (+34)941299621
E-mail: jesusmanuel.peregrina@unirioja.es
for other monosaccharides are lower still (e.g., glucose 9mꢀ1
;
[b] Dr. G. Joshi, Prof. A. P. Davis
xylose 5mꢀ1, galactose 2mꢀ1 and N-acetylgalactosamine 2mꢀ1).
This system presents an opportunity to study the recognition
of b-O-GlcNAc by a binding site that, though lectin-like, avoids
the size and complexity of a protein scaffold. Here we present
molecular dynamics and NMR spectroscopic investigations that
School of Chemistry, University of Bristol
Cantock’s Close, Bristol, BS8 1TS (UK)
[c] Prof. J. Jimꢂnez-Barbero
Centro de Investigaciones Biolꢃgicas (CSIC)
Ramiro Maeztu 9, 28040 Madrid (Spain)
Supporting information for this article is available on the WWW under
http://dx.doi.org/10.1002/cbic.201000526.
110
ꢁ 2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemBioChem 2011, 12, 110 – 117

Molecular Recognition
the carbonyl group NH2_C=O. Finally, an important hydrogen
bond was present between the b-glycosidic oxygen (O1) and
NH5. These hydrogen bonds are summarized in Figure 1B. On
the other hand, hydrophobic contacts were observed between
the methyl group of NHAc and the two spacer aromatic rings.
The averaged distance between the carbon atom of the
methyl group and the centers of the aromatic rings was calcu-
lated to be 3.6 Š. Notably, an additional hydrophobic contact
was detected between the methoxy group and one spacer ar-
omatic ring. As in the former cases, the distance between the
ring and the methyl group was 3.6 Š (Figure 1C). Experimental
evidence for this CH–p interaction[9] is the perturbation of the
chemical shift of the OMe group in the bound state with re-
spect to the free sugar. In addition, from a theoretical point of
view, when ligand 1 is removed from the complex, the side
chain attached to the spacer involved in the CH–p interactions
becomes significantly more flexible (see the Supporting Infor-
mation).
To check the reliability of the above-mentioned MD simula-
tions on these systems, we also ran 25 ns unrestrained MD
simulations on the receptor–1 complex in explicit water. Now
all the receptor atoms, including the water-solubilizing side
chains and the counter ions, were included in the calculations.
Strikingly, the distances derived from these simulations were
similar to the experimental ones. Moreover, the same interac-
tions between the receptor and compound 1 were detected,
which validates the unrestrained MD simulations.
Encouraged by these results, we decided to run unrestrained
MD simulations on the complexes formed between the recep-
tor and compounds 2 and 3. As before, the hydrogen bonds
observed in the complex with 1 were also found for these sub-
strates. However, in sharp contrast with 1, the population of
the hydrogen bond NH5···O1 dropped to only 20% in com-
pound 2 and, moreover, was never populated for 3 (with O1
axially oriented, Figure 2). Considering that the rest of the sta-
bilizing interactions have a similar weight in the three com-
plexes, and assuming that a typical NH···O hydrogen[10] bond
stabilizes around 1.9 kcalmolꢀ1, it could be deduced that the
Ka of compound 1 should be around 25 times the affinity con-
stant of derivative 3, which is in excellent agreement with the
experimental data commented on above. On the other hand,
the low population of this hydrogen bond obtained for the
receptor–3 complex indicates that this interaction, with the
proper orientation of NH5, plays the key role in the b versus a
selectivity of this receptor.
Scheme 1. A) Carbohydrates studied in this work. B) Structure of the recep-
tor used in this study, showing the labeling system used throughout the
text.
Taking into account the strong selectivity of the receptor for
compound 1 and the fact that b-O-GlcNAc appears in most
cases linked to a serine or a threonine residue; we decided to
use this simple lectin-like receptor to investigate the effect
that the underlying amino acid has on the binding properties.
On this basis, and given that the incorporation of unnatural
residues in the backbone of small glycopeptides can stabilize
conformations present in naturally occurring molecules or
exhibit some atypical conformations,[11] we report herein the
binding properties of different model glycopeptides derived
not only from natural serine and threonine (compounds 4 and
5, respectively) but also derived from the unnatural a-methyl-
Figure 1. A) Different orientations of the ensemble obtained from the MD-
tar simulations for complex receptor:1. The resulting structure reproduces
all the NOE-derived distances reported by some of us (ref. [6]). B) Schematic
representation of the different hydrogen bonds obtained from the MD-tar
simulations in the receptor–1 complex. C) Schematic representation of the
hydrophobic interactions between ligand 1 and the receptor. The hydrogen
atoms have been removed for clarity.
teracts with the carbonyl group NH1_C=O. On the other hand,
O6 of the sugar participates in two hydrogen bonds: O6···NH3
and O6···NH7. Moreover, O4 interacts with NH4 and O3ꢀH with
ChemBioChem 2011, 12, 110 – 117
ꢁ 2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chembiochem.org
111

J. M. Peregrina et al.
Scheme 2. Glycopeptides studied in this work, showing the la-
beling system used throughout the text.
Figure 2. Population of the hydrogen bond NH5···O1 between the receptor and carbohy-
drates 1, 2, and 3, as obtained from the 25 ns unrestrained MD simulations in explicit
water.
serine (MeSer) and a-methylthreonine (MeThr), compounds 6
and 7 (Scheme 2). Thus, the main goal of this study was to in-
vestigate in detail the role that the substituents at the Ca and/
or Cb atoms have in the glycopeptide conformations and, con-
sequently, on their binding properties. In these compounds,
the amino and carboxylic acid functional groups were trans-
formed into amides to simulate the peptide backbone.
Prior to carrying out the conformational study of the glyco-
peptides in the bound state, it is important to know the con-
formation of the glycopeptides in water (free state). In this
context, the conformational behavior of compounds 4 and 5
has been previously reported.[12]
In this case, the main conclusion is that although these com-
pounds have a similar behavior for the peptide backbone, rep-
resented by extended conformations, they have different be-
Figure 3. Ensembles obtained from the MD-tar simulations in explicit water
for model glycopeptides 6 and 7 together with detailed information about
their major conformations found in the free state for the most relevant tor-
sional angles.
havior in terms of the ys (C1-O1-Cb-Ca) dihedral angle. In fact,
for derivative 5, this angle showed values around 1208, result-
ing in an eclipsed conformation for the HbꢀCb and O1ꢀC1
bonds. This defined conformation can accommodate a water
pocket between the peptide and the carbohydrate moieties.
Additionally, the lateral chain adopts a g(+) conformation (c1 =
608) in both derivatives. On the other hand, the conformational
analysis of glycopeptides 6 and 7 was carried out by following
our methodology,[8] which combines NOE-derived distances
with MD-tar simulations. The results obtained for these two
novel molecules are summarized in Figure 3. Our data show
that the peptide backbone of both derivatives 6 and 7 adopt-
ed mainly helix-like conformations, which are typical for glyco-
peptides, which incorporate a,a-disubstitued amino acids. This
result is in good agreement with the weak-to-medium NOE ob-
served between the NH protons of the backbone (Supporting
Information). As far as the lateral chain (c1 torsion angle) is
concerned, the simulations suggest that the rotation around c1
in the MeThr derivative is restricted with values for this torsion
angle close to ꢀ608. In sharp contrast, the lateral chain of the
MeSer-containing glycopeptide is rather flexible, showing sig-
nificant population of each staggered conformation for this
torsion angle. Concerning the glycosidic linkage, they also
showed a different behavior in terms of the ys dihedral angle.
In fact, for 7, this angle showed values around 1208–1408. This
result is similar to that obtained for glycopeptide 5, which also
possesses a b-methyl group. However, ys was found to be
much more flexible for the MeSer analogue 6, which shows a
major anti-type arrangement for the GlcNAc residue and the
peptide moiety. Curiously, although the anomeric linkage and
the chemical nature of the sugar is completely different, it is
noteworthy that an eclipsed ys conformation has been previ-
ously reported by us for glycopeptides containing the a-O-
GalNAc-Thr unit.[13] This similar behavior could indicate that the
conformational tendencies for ys in O-glycopeptides are
mainly determined by the presence (or not) of the b-methyl
group in the b-hydroxy-a-amino acid residue (Thr or Ser, re-
spectively), irrespective of the chemical nature of the carbohy-
drate moiety and of the configuration at the anomeric center.
From the interactions perspective, the different features ob-
served in the glycopeptides could have important implications
for molecular recognition processes. To test this hypothesis, we
studied the binding of compounds 4–7 to the receptor in D2O
with 1H NMR spectroscopic titrations, following the methodolo-
gy previously described.[6]
As can be observed in Figure 4, when nine equivalents of
the glycopeptides 4–7 were added to a solution of the recep-
tor, the signals of this molecule became more complicated,
which is consistent with the loss of symmetry upon binding.
112
www.chembiochem.org
ꢁ 2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemBioChem 2011, 12, 110 – 117

Molecular Recognition
Figure 4. 1H NMR spectra of the receptor (0.5 mm) in D2O after addition of
nine equivalents of the different glycopeptides. Only aromatic signals of the
receptor are shown.
The association constants were measurable through the inte-
gral ratios of the bound and unbound receptor signals. The Ka
and DG values obtained for the different glycopeptides are
summarized in Table 1.
Figure 5. Ensembles obtained from the 25 ns unrestrained MD simulations
in explicit water for the complexes between the receptor and glycopeptides
4–7.
Table 1. Association constants (Ka) and DG values in aqueous solution for
the receptor shown in Scheme 1B with different substrates.
Name
Compound
Ka [mꢀ1
]
DG [kcalmolꢀ1
]
1
2
3
4
5
6
7
bGlcNAcOMe
GlcNAc
1
2
3
4
5
6
7
630[6]
156[6]
24[6]
91
13
75
3.82
2.99
1.88
2.67
1.52
2.56
2.27
aGlcNAcOMe
bGlcNAcSer
bGlcNAcThr
bGlcNAcMeSer
bGlcNAcMeThr
46
Important conclusions can be drawn from the inspection of
these Ka values. Firstly, the presence of the amino acid reduces
the affinity constant significantly. Secondly, the presence of a
and/or b-methyl groups at the amino acid moiety appears to
have negative effects on the binding. Especially relevant is the
small Ka value obtained for the Thr-containing glycopeptides.
In fact, the two glycopeptides with a b-methyl group (com-
pounds 5 and 7) exhibited the lower Ka values. According to
these values, the complex with glycopeptide 4 is, at 298 K,
about 1.1 kcalmolꢀ1 more stable than the complex with the
Thr derivative 5. Finally, it is important to note that the pattern
shown by the complex with ligand 4 (similar to that found for
compound 1) differs from that observed for the rest of the
glycopeptides, which could indicate that the methyl groups
accommodate the sugar moiety in a different way.
Figure 6. fp/yp distributions (backbone) obtained from the 25 ns unrestrain-
ed MD simulations in explicit water for the bound state of the glycopepti-
des.
To explain these results, we ran extensive MD simulations on
the complexes between glycopeptides 4–7 and the receptor.
Figure 5 shows 15 snapshots taken from these simulations for
each complex.
like conformation allows the formation of a weak hydrogen
bond between the amino group of the amino acid residues
and the carbonyl group NH1_C=O of the receptor.
As shown in Figure 6, the helix-like conformation was partic-
ularly populated in the bound state for glycopeptides 5 and 7
(compound with a b-methyl group); this conformer coexists
with the G conformation[14] in derivatives 4 and 6. The helix-
Concerning the glycosidic linkage (Figure 7), glycopeptides 4
and 6 adopted an alternate conformation, which is expected
ChemBioChem 2011, 12, 110 – 117
ꢁ 2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chembiochem.org
113

J. M. Peregrina et al.
formation of a hydrogen bond between the carbonyl group
NH2_C=O of the receptor and O3-H of the ligand. These two
features could explain the observed very low affinity of the
receptor for glycopeptides 5 and 7.
We guessed that the presence of the methyl groups in the
amino acid moiety could also modulate the ligand–receptor in-
teractions and, consequently, the affinity constant. As com-
mented on above, there is one hydrophobic contact (CH–p
interaction) that involves the OMe group of compound 1 and
the aromatic ring of one of the side chains (Figure 1C). On this
basis, Figure 9 shows the distances between the different car-
Figure 7. fs/ys distributions (glycosidic linkage) obtained from the 25 ns un-
restrained MD simulations in explicit water for the bound state of the glyco-
peptides.
for derivatives without a b-methyl group at the backbone.[12,13]
Moreover, glycopeptide 5 showed the typical eclipsed confor-
mation for the glycosidic linkage. Strikingly, and in contrast to
the result obtained for the free state, the alternate conforma-
tion was also the most populated for derivative 7 in the
bound state.
As a next step, the intermolecular hydrogen-bonding analy-
sis for all the complexes was carried out. The most persistent
hydrogen bonds between the sugar moiety and the receptor
are shown in Figure 8. The population of the hydrogen bond
NH5···O1 considerably differs from that observed for com-
pound 1; its occupancy was only around 25% in the best case
(compound 4). In particular, this hydrogen bond was very
weak (ꢁ5% occupancy) in compounds with a methyl group at
Cb (derivatives 5 and 7). This methyl group also inhibits the
Figure 9. A) Left panel) Superimposition of 10 conformers randomly taken
from the MD simulations for the complex with ligand 4, showing the CH–p
interaction of the backbone with the side chain of the receptor. Right panel)
Schematic representation of the distances between the different carbon
atoms and the center of the aromatic ring. B) Distances Cx–Ar obtained
from the unrestrained MD simulations for the different complexes.
bons of the backbone and the center of this aromatic ring. For
compounds 1 and 3, the distance between the OMe group
and the aromatic ring (Ar) is less than 3.6 Š. In compound 4,
without methyl groups in the backbone, the distance Cb–Ar is
also less than 3.6 Š. This scenario considerably differs when a
methyl group at a and/or b positions is present. The presence
of this additional moiety inhibits the CH–p interaction, with
distances Cb–Ar and Ca–Ar greater than 4.2 Š. Although the
additional methyl groups of compounds 5–7 could make con-
tact with the Ar group, these interactions should be rather
weak due to the large distance CꢀAr. In fact, only in the case
of compound 6 is this value less than 3.8 Š.
On the other hand, to compare our theoretical data to the
experimental Ka values, we calculated the DDG of binding for
compounds 4 and 5. To this end, we used the thermodynamic
integration (TI) methodology[15,16] as implemented in Amber 10
package (see the Supporting Information for computational
details). This method has been successfully employed in bio-
physical studies to predict free energy changes, notably the
Figure 8. Occupancy of the most significant hydrogen bonds obtained from
the 25 ns unrestrained MD simulations for the bound states of compounds
~
&
^
*
1 ( ), 2 ( ), 3 ( ), 4 (”), 5 (*), 6 ( ), and 7 (+).
114
www.chembiochem.org
ꢁ 2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemBioChem 2011, 12, 110 – 117

Molecular Recognition
effect of amino acid point mutations.[17] Based on these calcu-
lations, the difference in the DDGbinding was calculated to be
0.6 kcalmolꢀ1 in favor of compound 4, which is in the same di-
rection as the experimental data.
tacts, which may only take place for certain orientations of the
receptor groups and for the proper stereochemistries and sub-
stitutions in the glycopeptide chain.
Finally, to explain the high Ka value (>1000mꢀ1) observed
for the CKII-based O-GlcNAc decapeptide 8,[6] we ran 25 ns un-
restrained MD simulation in explicit water on the receptor–8
complex (Figure 10). In this case, the calculations were carried
Experimental Section
Synthesis: The synthesis of the receptor[6] and compounds 4 and
5[12] have been previously reported. The synthetic routes to obtain
glycopeptides 6 and 7 are described in the Supporting Informa-
tion.
Compound 6: 1H NMR (400 MHz, D2O): d=1.42 (s, 3H), 2.00 (s,
3H), 2.06 (s, 3H), 2.72 (s, 3H), 3.40–3.49 (m, 2H), 3.51–3.59 (m, 1H),
3.67–3.82 (m, 3H), 3.88–3.99 (m, 1H), 4.17 (d, J=9.9 Hz, 1H),
4.51 ppm (d, J=8.4 Hz, 1H); 13C NMR (100 MHz, D2O): d=19.8,
22.1, 22.2, 26.0, 55.5, 59.4, 60.7, 69.9, 71.7, 73.5, 75.8, 101.5, 173.7,
174.5, 174.7 ppm; elemental analysis calcd (%) for C15H27N3O8: C
47.74, H 7.21, N 11.13; found: C 47.81, H 7.22, N 11.10.
1
Compound 7: H NMR (400 MHz, D2O): d=1.00 (d, J=6.4 Hz, 3H),
1.31 (s, 3H), 1.94 (s, 3H), 1.99 (s, 3H), 2.65 (s, 3H), 3.37–3.47 (m,
2H), 3.48–3.55 (m, 1H), 3.61–3.69 (m, 1H), 3.76 (dd, J=12.1 Hz, J=
5.0 Hz, 1H), 3.86–3.99 (m, 2H), 4.51 ppm (d, J=8.3 Hz, 1H);
13C NMR (100 MHz, D2O): d=13.5, 14.4, 22.0, 22.3, 26.1, 55.3, 60.7,
62.9, 69.9, 73.4, 75.9, 78.5, 99.4, 173.1, 174.3, 174.7 ppm; elemental
analysis calcd (%) for C16H29N3O8: C 49.10, H 7.47, N 10.74; found: C
49.21, H 7.44, N 10.70.
Figure 10. Calculated ensembles obtained from the 25 ns unrestrained MD
simulations in explicit water for glycopeptide 8.
out without the externally directed, water-solubilizing side
chains. The MD simulations revealed that the larger backbone
not only allow further interactions between the peptide and
the receptor, but also increases the fraction of the key hydro-
gen bond NH5···O1 up to 60%, which could be at the origin of
the large experimentally measured Ka value. Although merely
speculative, the high affinity could be also attributed to the
electrostatic interactions between the positively charged back-
bone and the negatively charged side chains of the receptor.
The binding free energy for the receptor–8 complex was
estimated with the MM-PBSA method as implemented in
Amber 10 (see the Experimental Section for details). This bind-
ing energy (without considering the unfavorable entropy con-
tribution) was very high: about ꢀ40 kcalmolꢀ1. This result is
qualitatively in agreement with the experimental data.
NMR spectroscopy experiments: All the NMR spectroscopy ex-
periments were recorded on a Bruker Avance 400 spectrometer at
293 K. 1H and 13C NMR spectra were recorded in D2O (chemical
shifts are reported in ppm on the d scale). Magnitude-mode ge-2D
COSY spectra were recorded with gradients and by using the co-
sygpqf pulse program with 908 pulse width. Phase-sensitive ge-2D
HSQC spectra were recorded by using a z-filter and selection
before t1 by removing the decoupling during acquisition by use of
invigpndph pulse program with CNST2 (JHC)=145. 2D NOESY ex-
periments were made by using phase-sensitive ge-2D NOESY for
CDCl3 spectra and phase-sensitive ge-2D NOESY with WATERGATE
for H2O/D2O (9:1) spectra. Selective ge-1D NOESY experiments
were carried out by using the 1D-DPFGE NOE pulse sequence.
NOEs intensities were normalized with respect to the diagonal
peak at zero mixing time.
1H NMR spectroscopic titrations: Solutions of glycopeptides 4–7
were made up in D2O (99.9%). Aliquots were then added to an
NMR tube containing the receptor in the same solvent (500 mL, [re-
ceptor]initial =0.50 mm). The sample tube was shaken carefully after
each addition and 1H NMR spectra were recorded at 298 K on a
Brüker Avance 500 spectrometer. Affinity constants (Ka) were ob-
tained by integration of signals from both the free and bound re-
ceptor. Values were obtained from several spectra and the results
averaged. Details are given in the Supporting Information.
Conclusions
The recognition of b-O-GlcNAc-containing moieties by the syn-
thetic lectin is modulated by a variety of factors, which have
been scrutinized by using a combination of NMR spectroscopic
methods and modeling protocols. For more complex mole-
cules, additional interactions take place in the process. In this
particular case, and for the studied compounds, the presence
of a key hydrogen bond between the receptor and the OMe
group of compound 1 appears to be responsible for the high
selectivity observed for this compound. On the other hand, the
presence of the underlying amino acid residue decreases the
affinity constant, especially in those cases in which a b-methyl
group is present. Thus, the serine-containing glycopeptide ex-
hibits the highest affinity constant of the glycopeptides, and
the threonine derivative shows the lowest one. The combina-
tion of NMR spectroscopic and MD methods permits us to de-
duced that the observed selectivity has its origin in the forma-
tion of specific hydrogen bonds and hydrophobic (CH–p) con-
Molecular dynamics simulations
Unrestrained MD simulations in explicit water: Simulations were per-
formed by using the AMBER 10 program package[18] (parm03),[19]
which was implemented with GLYCAM 06 and GAFF parame-
ters[20,21] to accurately simulate the conformational behavior of the
sugar moiety and the receptor, respectively. The starting geome-
tries were generated by superimposing the carbohydrate moiety
of the glycopeptides on the structure previously proposed for the
receptor–1 complex.[6] Counterions (12Na+) and a cubic box of
pre-equilibrated TIP3P water molecules[22] were then added by
using Xleap module of AMBER. The simulations were run with the
ChemBioChem 2011, 12, 110 – 117
ꢁ 2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chembiochem.org
115

J. M. Peregrina et al.
PMEMD module of AMBER with SHAKE algorithm,[23] by using peri-
odic boundary conditions, a 2 fs time step, a temperature of 300 K,
a Langevin-type thermostat[24] for temperature control, and con-
stant pressure of 1 atm. 9 Š cut-off was applied to the Lennard–
Jones interactions, and Ewald sums for the treatment of the elec-
trostatic interactions.[25] An initial 2500 cycles of minimization (com-
bining steepest descent with conjugate gradient) were run by first
restraining the atoms of the complex (Na+ ions and water mole-
cules were allowed to move). The whole system was then mini-
mized by using 2000 cycles. This first step was followed by 200 ps
of dynamics at constant volume with weak positional restrains on
Spiro, Glycobiology 2002, 12, 43R–56R; e) C. Butkinaree, K. Park, G. W.
Hart, Biochim. Biophys. Acta Gen. Subj. 2010, 1800, 96–106; f) R. Hurta-
do-Guerrero, H. C. Dorfmueller, D. M. F. van Aalten, Curr. Opin. Struct.
Biol. 2008, 18, 551–557.
[2] G. W. Hart, M. P. Housley, C. Slawson, Nature 2007, 446, 1017–1022.
[3] F. I. Comer, G. W. Hart, J. Biol. Chem. 2000, 275, 29179–29182.
[4] N. E. Zachara, G. W. Hart, Biochim. Biophys. Acta Gen. Subj. 2004, 1673,
13–28.
[5] a) Y. Akimoto, L. K. Kreppel, H. Hirano, G. W. Hart, Diabetologia 2000, 43,
1239–1247; b) M. Brownlee, Nature 2001, 414, 813–820; c) K. Vosseller,
L. Wells, M. D. Lane, G. W. Hart, Proc. Natl. Acad. Sci. USA 2002, 99,
5313–5318.
[6] a) Y. Ferrand, M. P. Crump, A. P. Davis, Science 2007, 318, 619–622; b) Y.
Ferrand, E. Klein, N. P. Barwell, M. P. Crump, J. JimØnez-Barbero, C.
Vicent, G. J. Boons, S. Ingale, A. P. Davis, Angew. Chem. 2009, 121, 1807–
1811; Angew. Chem. Int. Ed. 2009, 48, 1775–1779.
[7] a) A. E. Torda, R. M. Scheek, W. F. van Gunsteren, Chem. Phys. Lett. 1989,
157, 289–294; b) D. A. Pearlman, P. A. Kollman, J. Mol. Biol. 1991, 220,
457–479; c) A. E. Torda, R. M. Scheek, W. F. van Gunsteren, J. Mol. Biol.
1990, 214, 223–235; d) A. E. Torda, R. M. Brunne, T. Huber, H. Kessler,
W. F. van Gunsteren, J. Biomol. NMR 1993, 3, 55–66; e) D. A. Pearlman,
J. Biomol. NMR 1994, 4, 279–299; f) A. P. Nanzer, W. F. van Gunsteren,
A. E. Torda, J. Biomol. NMR 1995, 6, 313–320.
[8] a) F. Corzana, J. H. Busto, S. B. Engelsen, J. JimØnez-Barbero, J. L. Asensio,
J. M. Peregrina, A. Avenoza, Chem. Eur. J. 2006, 12, 7864–7871; b) F.
Corzana, J. H. Busto, G. JimØnez-OsØs, J. L. Asensio, J. JimØnez-Barbero,
J. M. Peregrina, A. Avenoza, J. Am. Chem. Soc. 2006, 128, 14640–14648.
[9] a) S. Vandenbussche, D. Diaz, M. C. Fernꢀndez-Alonso, W. Pan, S. P. Vin-
cent, G. Cuevas, F. J. CaÇada, J. JimØnez-Barbero, K. Bartik, Chem. Eur. J.
2008, 14, 7570–7578; b) G. Terraneo, D. Potenza, A. Canales, J. JimØnez-
Barbero, K. K. Baldridge, A. Bernardi, J. Am. Chem. Soc. 2007, 129, 2890–
2900; c) K. Ramírez-Gualito, R. Alonso-Ríos, B. Quiroz-García, A. Rojas-
Aguilar, D. Díaz, J. JimØnez-Barbero, G. Cuevas, J. Am. Chem. Soc. 2009,
131, 18129–18138.
the complex (10 kcalmolꢀ1 ꢀ2). In this step the system was heated
Š
form 100 to 300 K. The restrains on the solute were then removed
and we ran then 400 ps MD simulations at 300 K and 1 atm to get
the appropriate density. Finally, we ran a 25 ns MD simulations by
using the conditions commented on above.
MD-tar simulations: They were performed with AMBER 6.0 pack-
age[26] (parm03),[19] which was implemented with GLYCAM 06 pa-
rameters[20] to accurately simulate the conformational behavior of
the sugar moiety. NOE-derived distances were included as time-
averaged distance constraints. A hrꢀ6iꢀ1/6 average was used for the
distances. Final trajectories were run by using an exponential
decay constant of 8000 ps and a simulation length of 80 ns for the
MD-tar simulation with e=80 to simulate the water environment.
Thermodynamic integration method: These simulations were carried
out as implemented in AMBER 10[8] by using the default parame-
ters. In our calculations compound 5 was transformed into deriva-
tive 4. The non-TI parts of the input files were carried out by using
the following steps: Firstly, 500 steps of steepest-descent minimiza-
tion, secondly, 500 ps of density equilibration and finally 200 ps of
NTP production MD to collect dV/dl data. A time step of 2 fs was
used together with the Shake algorithm, a 9 Š direct sum cutoff,
isotropic pressure scaling, and a Langevin-type thermostat. The l
parameter was set to 0.1, 0.2,… 0.9. Only for step 2, a soft-core po-
tential were activate (by setting ifsc=1). This uses a different form
of the Lennard–Jones equation, specifically designed for better
convergence of TI calculations in the case of appearing or disap-
pearing atoms. Plots showing dV/dl as function of l are shown in
the Supporting Information.
[10] J. Emsley, Chem. Soc. Rev. 1980, 9, 91–124.
[11] a) A. Fernꢀndez-Tejada, F. Corzana, J. H. Busto, G. JimØnez-OsØs, J. M.
Peregrina, A. Avenoza, Chem. Eur. J. 2008, 14, 7042–7058; b) A. Fernꢀn-
dez-Tejada, F. Corzana, J. H. Busto, A. Avenoza, J. M. Peregrina, Org.
Biomol. Chem. 2009, 7, 2885–2893; c) A. Fernꢀndez-Tejada, F. Corzana,
J. H. Busto, A. Avenoza, J. M. Peregrina, J. Org. Chem. 2009, 74, 9305–
9313.
[12] a) A. Fernꢀndez-Tejada, F. Corzana, J. H. Busto, G. JimØnez-Oses, J. JimØ-
nez-Barbero, A. Avenoza, J. M. Peregrina, Chem. Eur. J. 2009, 15, 7297–
7301; b) C. Aydillo, G. JimØnez-OsØs, J. H. Busto, J. M. Peregrina, A. Ave-
noza, M. M. Zurbano, Chem. Eur. J. 2007, 13, 4840–4848.
[13] F. Corzana, J. H. Busto, G. JimØnez-OsØs, M. García de Luis, J. L. Asensio,
J. JimØnez-Barbero, J. M. Peregrina, A. Avenoza, J. Am. Chem. Soc. 2007,
129, 9458–9467.
MM-PBSA method:[27] The simulation to estimate the binding
energy between the receptor and glycopeptide 8 was carried out
as implemented in Amber 10 program package,[18] by using the
default keywords in the MMPBSA.py script (http://ambermd.org/
tutorials/advanced/tutorial3/py_script/files/mmpbsa_py.tar.gz).
[14] Letter code employed in S. S. Zimmerman, M. S. Pottle, G. Nemethy,
H. A. Scheraga, Macromolecules 1977, 10, 1–9.
[15] a) T. Simonson in Computational Biochemistry and Biophysics (Eds.: O.
Becker, A. D. MacKerell, B. Roux, M. Watanabe) Marcel Dekker, New York,
2001; b) P. Kollman, Chem. Rev. 1993, 93, 2395–2417; c) M. L. Lamb,
W. L. Jorgensen, Curr. Opin. Chem. Biol. 1997, 1, 449–457.
[16] T. Steinbrecher, D. A. Case, A. Labahn, J. Med. Chem. 2006, 49, 1837–
1844.
[17] T. Simonson, G. Archontis, M. Karplus, J. Phys. Chem. B 1997, 101, 8349–
8362.
[18] P. A. Kollman et al, AMBER 10, University of California, San Francisco,
2008.
Acknowledgements
We are grateful to the Ministerio de Ciencia e Innovaciꢃn
(CTQ2009-13814/BQU and grant to A.F.-T.) the Universidad de La
Rioja (EGI10/65), the Gobierno de La Rioja (Colabora 2010/05),
the EPSRC (research grant EP/D060192/1) and the Common-
wealth Scholarship Commission (Scholarship to G.J.). We also
thank CESGA for computer support.
[19] Y. Duan, C. Wu, S. Chowdhury, M. C. Lee, G. Xiong, W. Zhang, R. Yang, P.
Cieplak, R. Luo, T. Lee, J. Caldwell, J. Wang, P. A. Kollman, J. Comput.
Chem. 2003, 24, 1999–2012.
[20] K. N. Kirschner, A. B. Yongye, S. M. Tschampel, J. Gonzꢀlez-Outeirino,
C. R. Daniels, B. L. Foley, R. J. Woods, J. Comput. Chem. 2008, 29, 622–
655.
[21] J. Wang, R. M. Wolf, J. W. Caldwell, P. A. Kollman, D. A. Case, J. Comput.
Chem. 2004, 25, 1157–1174.
Keywords: conformational analysis · glycopeptides · molecular
dynamics · molecular recognition · NMR spectroscopy
[1] a) L. Wells, K. Vosseller, G. W. Hart, Science 2001, 291, 2376–2378; b) J. E.
Rexach, P. M. Clark, L. C. Hsieh-Wilson, Nat. Chem. Biol. 2008, 4, 97–106;
c) N. E. Zachara, G. W. Hart, Chem. Rev. 2002, 102, 431–438; d) R. G.
[22] M. W. Mahoney, W. L. Jorgensen, J. Chem. Phys. 2000, 112, 8910–8922.
[23] J.-P. Ryckaert, G. Ciccotti, H. J. C. Berendsen, J. Comput. Phys. 1977, 23,
327–341.
116
www.chembiochem.org
ꢁ 2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemBioChem 2011, 12, 110 – 117

Molecular Recognition
[24] S. A. Adelman, J. D. Doll, J. Chem. Phys. 1976, 64, 2375–2388.
[25] T. Darten, H. Lee, L. G. Pedersen, J. Chem. Phys. 1995, 103, 8577–8593.
[26] P. A. Kollman et al., AMBER 6, University of California, San Francisco,
1999.
[27] J. Wang, P. Morin, W. Wang, P. A. Kollman, J. Am. Chem. Soc. 2001, 123,
5221–5230.
Received: September 3, 2010
Published online on December 22, 2010
ChemBioChem 2011, 12, 110 – 117
ꢁ 2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chembiochem.org
117