Lanthanide-chelating carbohydrate conjugates are useful tools to characterize carbohydrate conformation in solution and sensitive sensors to detect carbohydrate-protein interactions

Article abstract of DOI:10.1021/ja502406x

The increasing interest in the functional versatility of glycan epitopes in cellular glycoconjugates calls for developing sensitive methods to define carbohydrate conformation in solution and to characterize protein-carbohydrate interactions. Measurements of pseudocontact shifts in the presence of a paramagnetic cation can provide such information. In this work, the energetically privileged conformation of a disaccharide (lactose as test case) was experimentally inferred by using a synthetic carbohydrate conjugate bearing a lanthanide binding tag. In addition, the binding of lactose to a biomedically relevant receptor (the human adhesion/growth-regulatory lectin galectin-3) and its consequences in structural terms were defined, using Dy³⁺, Tb³⁺, and Tm³⁺. The described approach, complementing the previously tested protein tagging as a way to exploit paramagnetism, enables to detect binding, even weak interactions, and to characterize in detail topological aspects useful for physiological ligands and mimetics in drug design.

Full text of DOI:10.1021/ja502406x

Article
pubs.acs.org/JACS
Lanthanide-Chelating Carbohydrate Conjugates Are Useful Tools To
Characterize Carbohydrate Conformation in Solution and Sensitive
Sensors to Detect Carbohydrate−Protein Interactions
†
‡
‡
∥
́
́
́
Angeles Canales, Alvaro Mallagaray, M. Alvaro Berbís, Armando Navarro-Vazquez,
́
Gema Domínguez,§ F. Javier Canada,‡ Sabine Andre,# Hans-Joachim Gabius,# Javier Perez-Castells,§
́
́
̃
and Jesus Jimenez-Barbero*,‡
́
́
†Department of Química Organ
Madrid, Spain
́
ica I, Fac. C.C. Químicas, Universidad Complutense de Madrid, Avd. Complutense s/n 28040,
‡Department of De Biología Físico-Química, Centro de Investigaciones Biolog
́
icas, CSIC, 28040, Madrid, Spain
#Institute of Physiological Chemistry, Faculty of Veterinary Medicine, Ludwig-Maximilians University, 80539, Munich, Germany
§Faculty of De Farmacia, Department of Química, Universidad CEU San Pablo, Urb. Montepríncipe, ctra. Boadilla del Monte, 28668,
Madrid, Spain
∥Department of Química Organ
́
ica, Universidade de Vigo, Campus Universitario Vigo, 36310, Pontevedra, Spain
S
* Supporting Information
ABSTRACT: The increasing interest in the functional versatility
of glycan epitopes in cellular glycoconjugates calls for developing
sensitive methods to define carbohydrate conformation in solution
and to characterize protein−carbohydrate interactions. Measure-
ments of pseudocontact shifts in the presence of a paramagnetic
cation can provide such information. In this work, the energetically
privileged conformation of a disaccharide (lactose as test case) was
experimentally inferred by using a synthetic carbohydrate conjugate
bearing a lanthanide binding tag. In addition, the binding of lactose
to a biomedically relevant receptor (the human adhesion/growth-
regulatory lectin galectin-3) and its consequences in structural
terms were defined, using Dy3+, Tb3+, and Tm3+. The described approach, complementing the previously tested protein tagging
as a way to exploit paramagnetism, enables to detect binding, even weak interactions, and to characterize in detail topological
aspects useful for physiological ligands and mimetics in drug design.
the carbohydrate ligand5 and from the receptor.6,7 A widely
INTRODUCTION
■
popular strategy takes advantage of J-couplings or magnet-
ization transfer processes, related to intra- or intermolecular
nuclear Overhauser (NOE) enhancements,8 although the use
of other sources of information, for instance, residual dipolar
couplings (RDCs), has also been described.9,10 The growing
insight into physiological glycan functionality via lectin binding
is a strong incentive to broaden the scope of respective
analytical procedures. When planning to develop new methods,
the special features of carbohydrates as ligands deserve to be
considered. Carbohydrates are flexible molecules that can adopt
a variety of shapes in solution and bind in a certain geometry to
their particular receptors.11 Their binding strengths to proteins
are usually rather weak, unless multivalent effects are involved.12
In most cases, despite an exquisite selectivity of protein−
carbohydrate interactions, dissociation constants are typically
in the mM to low μM range.13 Also of note, there are often
Nature has endowed carbohydrates with exceptional properties
for coding information, and indeed, a sophisticated enzymatic
machinery has been developed for highly regulated biosynthesis
of complex glycans.1 This also holds true for receptors (lectins),
which translate the sugar-based signals into cellular effects.1,2
Addressing the challenge of structural characterization of
protein−carbohydrate recognition in detail will not only give
a functional meaning to distinct aspects of the glycome but
will also be the basis to develop potent inhibitors, e.g., to block
bacterial/viral infections or dysfunctional growth regulation in
inflammation/tumor progression. This biomedical perspective
explains the attention given to developing reliable and sensitive
methods to explore carbohydrate structure, dynamics, and
recognition.
Solution NMR spectroscopy is a major tool to characterize
these properties. In fact, different NMR approaches are widely
employed to study the interactions of carbohydrates and
chemical analogues (glycomimetics)3,4 with their receptors up
to the level of atomic resolution, both from the perspective of
Received: March 13, 2014
Published: May 15, 2014
© 2014 American Chemical Society
8011
dx.doi.org/10.1021/ja502406x | J. Am. Chem. Soc. 2014, 136, 8011−8017

Journal of the American Chemical Society
Article
problems in obtaining structural information on these systems
by NOEs, due to signal overlapping, strong coupling, and/or
the scarcity of key NOE information.14 Moreover, few inter-
molecular carbohydrate−protein NOEs are usually detected,
hampering the structure calculation of carbohydrate−protein
complexes at high resolution.11 Therefore, the NMR-based
structural characterization at atomic resolution of protein−
carbohydrate complexes is not a trivial task. As consequences,
from the NMR perspective, there is an obvious need for
introducing additional NMR parameters that can be reliably
employed to monitor and study the conformation of complex
carbohydrates and their interactions with receptors. Para-
magnetic ions offer this potential. Historically, the presence of
Mn2+ as an integral part in a plant lectin, which is essential for
activity, served as starting point for measuring magnetic
relaxation dispersion or line broadening.15
The use of paramagnetic ions exploiting lanthanide binding
tags (LBT) to obtain new NMR parameters providing
structural information has been extended in the past few
years for protein structure determination.16−18 The access to
pseudocontact shifts (PCS), paramagnetic relaxation enhance-
ments (PREs), and even to residual dipolar couplings (RDC)19
has permitted to deduce key conformational and dynamic
information in different elegant examples. In the study of
protein−carbohydrate interactions, Prestegard and co-workers
have made relevant contributions employing either spin-labeled
sugar molecules20 or measuring RDC parameters in chemically
modified proteins, either with short alkyl chains,21 histidine
tags,22 or lanthanide-binding peptides.23 Using a human lectin,
known for its role in infection, inflammation, pro-tumoral pro-
cesses, and glycoprotein routing, i.e. galectin-3,24−29 genetic
engineering led to a construct of lectin and chelating peptide
at the C-terminus.23 This hybrid combined capacity for sugar
binding with topological fixation of the lanthanide.
In a molecular recognition context, from the receptor’s
perspective, a paramagnetic protein tagging approach has been
recently implemented for fragment-based ligand screening.30
In comparison, few examples have taken the alternative
approach, i.e., tackling the problem from the ligand’s
perspective. We31 and others32,33 have recently reported on
the use of a chemically well-defined lanthanide binding tag,
to perform structural studies of carbohydrates in the free state,
with the final aim of also performing molecular recognition
studies. Our initial design31 consisted of a 1-β-aminochitobiose
moiety bearing a rigid linker attached to the 1-amino group,
with a biphenylic EDTA lanthanide-chelating unit. The 7.1 Å
length of the biphenyl moiety positions the metal ion more
than 10 Å away from the sugar moiety in the conjugate, thus
minimizing PRE in the NMR-active carbohydrate atoms. This
long distance to the metal center causes the observed chemical
shift changes that would be primarily due to the pseudocontact,
rather than contact shift mechanism. Very satisfactory results
were obtained, and the measurement and analysis of the
PCSs validated the conformational behavior of N,N′-diacetyl
chitobiose, the disaccharide repeating unit of chitin. This
example also allowed to demonstrate that the attached LBT at
the reducing end does not modify the shape of the disaccharide
entity. However, from the chemical perspective, this compound
presents a stereogenic center at the chelating unit, which
complicates the synthetic process.
Equally important, we also provide a proof-of-principle example
that these derivatives can be efficiently used to monitor
carbohydrate−protein interactions, providing structural infor-
mation on the binding event at the protein receptor. Of note,
depending on the choice of the metal ion at the tag, the magni-
tude of the paramagnetic effects can be modulated, facilitating
access to a variety of structural data. To enable compari-
son between methods, we have deliberately tested human
galectin-3, also offering a biomedically relevant study object.
In addition, this approach does not require the preparation of
protein mutants, thus allowing the study of the protein in its
native form.
METHODS
■
PCSs Measurements. Experimental PCSs for the lanthanide-
chelating carbohydrate derivative were calculated as the difference
between the chemical shift of each signal in a 1H−13C HSQC
spectrum acquired in the presence of a diamagnetic metal ion (La3+)
and a spectrum acquired in the presence of a paramagnetic metal ions
(Tb3+, Dy3+). Spectra were acquired with 16 scans, 2000 and 256
complex points in H and 13C dimensions, respectively, at 11.7 T
1
and 298 K. Samples were prepared in deuterated Tris buffer (D2O,
hexadeuterated Tris, 50 mM, pH 7.8) at a final concentration of
the carbohydrate derivative 5 of 2.5 mM. CH3CN (2 μL) was added as
chemical shift reference.
Once the experimental values were obtained, a development version
of Mspin software34 was used to back-calculate the expected PCSs
from the different lactose conformations. The existence of one major
and two minor conformations around the glycosidic angles of lactose
in solution has been previously described in detail.35−39 An ensemble
of lactose conformations was built taking into account both the
described conformers around the glycosidic linkages and the flexible
torsions of the linker.32,35−39 Three conformational families were built
for the lactose conjugate according to the Φ and Ψ dihedral angles
values, namely syn Φ/syn Ψ, anti Φ/syn Ψ, and syn Φ/anti Ψ (see
below) . The geometry of the phenylene diamino tetraacetic chelating
unit was obtained from the reported X-ray coordinates for this
moiety.40 The quality of the fitting between experimental and back-
calculated values is given by the Cornilescu’s Q factor, which is defined
by the following expression:34
2
∑ (PCScalc − PCSexp
)
Q =
2
∑ PCSexp
For molecular recognition applications, the ability of our lanthanide-
chelating lactose-derivative to be specifically recognized by Gal-3 was
studied by saturation transfer difference (STD),41 an NMR technique
that is based on the transfer of magnetization from the irradiated
protein to its binding ligands and provides epitope information at the
ligand level. Samples for STD experiments contained 1 mM of the
lanthanide chelating lactose-conjugate and 30 μM of human gal-3
CRD in deuterated Tris buffer, pH 7.8. NMR data were collected at
298 K on a Bruker Avance DRX 500 MHz spectrometer equipped
with a 5 mm inverse probe head. The protein was saturated on-
resonance at 0.5 ppm and off-resonance at 100 ppm with a train of
Gaussian-shaped pulses of 50 ms each and a total saturation time of 2 s.
A spin-lock field of 5 kHz with a duration of 15 ms was applied to
minimize the protein signal background.
1
Galectin-3 PCSs were measured from H−15N HSQC experiments
acquired in the presence of the lanthanide chelating carbohydrate
conjugate loaded either with diamagnetic (La3+) or with paramagnetic
ions (Tm3+, Dy3+). The spectra were recorded with 48 scans, 1024 and
128 complex points in 1H and 15N dimensions, respectively, at 14.1 T
and 298 K. Samples were prepared in deuterated Tris buffer (D2O,
hexadeuterated Tris, 50 mM, pH 7.8). The protein concentration was
270 μM, and the lactose derivative/protein molar ratio was set to 12/1.
For analyzing the lectin PCSs, MSpin software was also used to
back-calculate the expected PCSs for the geometry reported in the
Herein, we present the synthesis of a second generation of
carbohydrate derivatives with a novel lanthanide-binding tag
and its application for conformational analysis of carbohydrates.
8012
dx.doi.org/10.1021/ja502406x | J. Am. Chem. Soc. 2014, 136, 8011−8017

Journal of the American Chemical Society
Article
X-ray crystallographic structure of the galectin-3/lactose complex
(PDB code 2NN8). The lactose molecule of the X-ray structure was
replaced by the lactose conjugate bearing the lanthanide-binding tag,
and an ensemble of lactose structures was built, as described above, to
consider the flexibility of the linker.42 The geometry of the phenylene
diamino tetraacetic unit was taken from the reported coordinates of
this moiety crystallized in the presence of Fe(III).40
Scheme 1. Synthesis of the PhDTA Chelating Unit
Selectively Brominated
RESULTS AND DISCUSSION
■
Synthesis of the Lanthanide Chelating Carbohydrate
Conjugate. This second-generation lanthanide-carbohydrate
conjugate is the straightforward extension of the phenylene
diamino tetraacetic unit (PhDTA), which has been recently
used as chelating unit for caged lanthanide complexes in NMR
studies.42 Although shorter than our initial design, this structure
positions the metal at more than 10 Å from the carbohydrate
moiety, further away than the monoaromatic precursor. This
length permits to obtain PCS values from both sugar rings of
the disaccharide, as will be described below.
Scheme 2. Synthetic Scheme of the Boronation-Suzuki
Coupling Leading to the Lactose Tag
The synthesis of the lactose derivative 5 was planned in a
convergent way. 1-β-Aminolactose was obtained by synthesiz-
ing hepta-O-acetylated 1-β-aminolactose from 1-β-azide lactose
followed by deacetylation, thus obtaining the desired
disaccharide in a 75% yield.32 The synthesis of the linker was
based on a Suzuki coupling of an aryl boronate with protected
p-iodobenzoic acid. We chose trimethylsilylethyl (TMSE) as
protecting group for the carboxylic acid because it can be
selectively cleaved in the presence of tert-butyl esters and is
highly stable under Suzuki conditions.
First, we prepared the PhDTA fragment. Initially, we used
4-bromophenylenediamine as starting material, which upon
extensive N-alkylation with tert-butylbromoacetate gave the
trialkylated derivative 1 as the major reaction product. No
reaction conditions afforded the desired tetraacetate derivative.
Therefore, we decided to start with o-phenylenediamine
and to prepare compound 2 followed by a bromination
reaction.
This intermediate 2 was prepared (83% yield) under similar
alkylation conditions, with traces of other alkylation sub-
products, which could be separated. The use of acetonitrile as
solvent was critical as the alkylation of phenylenediamine in
toluene or DMF provided only low yields of mixtures of mono-,
di-, and trialkylated products. Bromination of 2 (Scheme 1)
with KBr/Oxone gave 3 as a single regioisomer (78% yield).
The next step was a tandem boronation-Suzuki coupling,
which, after an extensive study of the reaction conditions (see
Supporting Information), afforded the bicyclic product 4 in a
75% overall yield. The TMSE group was selectively deprotected
in the presence of TBAF affording the corresponding acid
in 86% yield, which was condensed with 1-β-aminolactose in
the presence of HATU and DIPEA. The final carbohydrate
derivative 5 was obtained upon hydrolysis of the tert-butylesters.
All the condensations and hydrolysis steps proceeded with good
yields (Scheme 2).
described in the Methods section. Fittingly, only one set of
NMR signals was observed for the molecule in the presence of
stoichiometric amounts of the metal. Inspection of the cross
peaks permitted to demonstrate that all the carbohydrate NMR
signals could be detected. Therefore, this linker is long enough
to restrict PREs effects to the linker NMR signals.
Assessing the Conformation of Lactose by PCSs. The
first step in the analysis concerned the employment of PCSs to
assess the conformation of the lactose-containing molecule in
water solution and to demonstrate that the shape of the
disaccharide is not affected by the presence of the LBT. Thus,
PCSs of the lactose derivative 5 were obtained by comparing
Significant induced 1H and 13C PCS could be observed in the
sugar even above 1 ppm for Glc H-1 (the sugar hydrogen
closest to the metal, ΔPCS = +1.78 ppm for Dy3+ and ΔPCS =
+0.94 ppm for Tb3+, see Figure 1).
1
the 13C and H NMR chemical shifts of 5 in the presence
of diamagnetic (La3+) and paramagnetic (Tb3+, Dy3+) ions,
1
through the use of standard H−13C HSQC experiments, as
8013
dx.doi.org/10.1021/ja502406x | J. Am. Chem. Soc. 2014, 136, 8011−8017

Journal of the American Chemical Society
Article
1
Figure 2. H and 13C PCSs correlation plots (b) for the fitting of the
experimental data to the back calculated data for the geometry of 5
loaded with Dy3+, built from the major lactose conformation (a). Eight
structures were considered for each conformer (synΦ/synΨ, antiΦ/
synΨ, or synΦ/antiΨ) in the calculations, two staggered conforma-
tions of the amide respect to the aromatic ring attached and four
conformations around the bond connecting both aromatic rings of the
biphenyl were used.
in a 0.064−0.068 range, see Figure 2. The value for the synΦ/
antiΨ conformations was markedly deviated (Q = 0.108−0.139),
while the antiΦ/synΨ conformers also provided a poorer fitting
(Q = 0.079−0.086).
1
Figure 1. Superimposition of the H−13C HSQC spectra obtained
for 5 acquired in the presence of a diamagnetic metal (La3+, red) and
in the presence of a paramagnetic metal (Tb3+, blue). The PCS
(0.94 ppm) or Glc H-1 is highlighted.
It is worth noting that since metal-proton vectors span a
narrow angle, the SVD problem is not well-conditioned, with
the corresponding condition numbers in the order of 100. This
means that the Δχ tensor can not be fully determined from
the available experimental data. However, since the metal-nuclei
vectors span similar angles in the three conformations, dis-
crimination between them is possible even when some elements
of the Δχ tensor are not well-defined. A bootstrapping analysis
of the Dy3+ data,44 assuming a Gaussian error of 0.01 ppm in the
experimental measurements, was then carried out in order to
check if the Q factor difference obtained between the three
conformational families showed statistical significance. It was
observed that the ΔQ difference between quality factors was
kept nearly constant (σ ∼ 0.001) among the different samples.
Hence, our PCS results agree with existence of one very
predominant conformation around the synΦ/synΨ glycosidic
linkages of lactose as stated before.35−39 This conclusion, which
has been deduced by using a variety of experimental and
theoretical methods,38,39 differs from that recently reported,32
using a similar approach to that described herein. In that work,
the authors indicated that it was necessary to include several
lactose conformations to fit the experimental data.
The experimental PCSs were then used to infer the solution
conformation of the lactose moiety around the glycosidic
torsion angles and to ascertain the presence of the typical chair
conformation for the two pyranose rings also in the pre-
sence of the LBT. Different studies have concluded that
the conformation of lactose can be described by two 4C1 chairs
for the Gal and Glc rings along with a major conformation
around the glycosidic torsion angles, Φ (H1−C1−O−C4) and
Ψ (C1−O−C4−H4).35−39 This conformation (synΦ/synΨ),
favored by the exoanomeric effect,43 is described by Φ/Ψ
torsion angle values of ∼55°/0°. Two other very minor
conformations have been detected, with proportions smaller
than 5%, with torsion angle values around 180° (anti) either
for Φ or for Ψ, thus establishing antiΦ/synΨ or synΦ/antiΨ
conformers, respectively.35−39
MSpin calculations34 were performed to deduce the expected
PCSs, using as input the coordinates of the three basic
geometries of the lactose conformers mentioned above and
the X-ray coordinates for the LBT40 in the presence of Tb3+
and Dy3+. Moreover, in order to take into account the linker
flexibility, four different staggered conformations of the
biphenyl moiety and two staggered positions of the amide
bond with respect to the adjacent aromatic ring were built.
Therefore, eight structures were calculated for each Φ/Ψ
conformer. The backcalculated PCSs were fairly similar for all
the rotamers since, as expected from the design, the rotation
around these bonds did not substantially alter the metal
position with respect to the sugar moieties. A representation of
the flexibility of the linker is given in Figure 2a for the major
synΦ/synΨ conformer.
Molecular Recognition: The Interaction of 5 with the
Carbohydrate Recognition Domain of Human Galectin-3
(hGal-3). It is well-established that molecular recognition
events in solution can be followed by using NMR, from either
the ligand’s or the receptor’s perspectives.45
We first tested the ability of 5 to be specifically recognized by
the human Gal-3 carbohydrate recognition domain (hGal3
CRD). As a first step, titration experiments were carried out
both with natural lactose (as blank experiment) and with the
lactose derivative 5 for comparison. The dissociation constants
obtained with both ligands are equivalent (see Supporting
Information).
Fittingly, the back-calculated PCSs obtained with the major
conformer (synΦ/synΨ) were in excellent agreement with the
experimental data (see Figure 2). The Q quality factor obtained
from the MSpin calculations is 0.06 for Dy3+, considering both
In addition, we used STD-NMR.41 STD experiments re-
present a straightforward means to explore, at atomic resolu-
tion, that the ligand epitopes responsible for the recognition
process by hGal3 were the same.
1
the H and 13C NMR data.
SVD-based computations (see Supporting Information for
details) were also performed using the MSpin software34 to fit
the observed 1H and 13C PCSs. As input coordinates, the three
conformational families of the tagged lactose conformers
mentioned above and the X-ray coordinates of the LBT40
were employed. In the presence of Tb3+ and Dy3+, the major
conformer geometry (synΦ/synΨ) furnished Q quality factors
The rise of STD signals in the corresponding experiment
demonstrated the recognition of 5 by the lectin (Figure 4).
In addition, inspection of the STD profile revealed a higher
degree of magnetization for the hydrogen nuclei belonging to
the lactose moiety (Figure 4B), leaving the linker and chelating
region comparatively much less affected. The highest degree of
8014
dx.doi.org/10.1021/ja502406x | J. Am. Chem. Soc. 2014, 136, 8011−8017

Journal of the American Chemical Society
Article
Regarding the PCS data, the interaction of 5 with hGal-3
CRD was studied from the point of view of the receptor, using
the observed chemical shift perturbation data of the protein
signals, obtained through 1H−15N HSQC experiments.46 Thus,
the 15N-labeled hGal-3 CRD was used for monitoring the
binding process. It was expected that the usual chemical shift
1
perturbation data of the H−15N cross-peaks of hGal-3 in the
presence of lactose would be further enhanced in the presence
of 5, due to the presence of the paramagnetic metal ion at the
LBT moiety. First, a blank experiment with 5 loaded with La3+
was performed, indicating that the chemical shift perturbations
induced in the hGal-3 signals were basically identical to those
obtained in the presence of just lactose, without the LBT
moiety (Figure 3). Thus, the binding modes of 5 and lactose to
hGal-3 are equivalent. Then, different protein samples were
titrated with increasing amounts of the lactose derivative 5 pre-
viously loaded with the chosen paramagnetic ions (Figure 5),
Figure 3. Right panels: 1H−15N HSQC (acquired at 700 MHz) for the
galectin-3:lactose (molar ratio 1:12) sample. Left panels: 1H−15N
HSQC (acquired at 600 MHz) for the sample containing galectin-3
and the lactose derivative 5, loaded with La3+ (molar ratio 1:12).
1
Figure 5. (a) Superimposition of H−15N HSQC spectra of hGal-3-
CRD in complex with 12 equiv of the lactose-analogue 5 acquired in
the presence of a diamagnetic metal (red, loaded with La3+) and in the
presence of a paramagnetic metal (blue, loaded with Tm3+). Some of
the residues with the highest PCSs are labeled. (b) Structure of h-gal3/
lactose derivative 5 complex built from the coordinates of PDB 2NN8.
Eight structures of the synΦ/synΨ conformer were considered in the
calculation to take into account the flexibility of the biphenyl linker.
(c) Comparison of the back-calculated PCS (MSpin calculation) for
the structure represented in (b) with those experimentally observed.
Figure 4. Lower panel: STD experiments performed on either 5 or
lactose in the presence of hGal-3 CRD. The sample contains (a) 5,
with labels (off-resonance experiment), (b) hGal3 CRD and 5 (STD
experiment), and (c) hGal3 CRD and lactose (STD experiment).
Upper panel: ligand epitope mapping of the recognition of 5 by
human hGal-3 CRD as inferred from STD experiments. Protons from
the PhDTA linker (labeled with asterisks, i.e., aromatic protons 9−17
and those of the chelating arms 18−20) give rise to a much weaker
STD response than those from the lactose moiety, thus indicating that
the recognition takes place at the level of the sugar.
as described in the Methods section. In detail, for these
experiments, two different paramagnetic metals were tested,
namely Dy3+or Tm3+.
In both cases, clear PCSs were observed for many protein
signals, demonstrating the transfer of the paramagnetic effects
to the protein nuclei from the noncovalently bound ligand and
thus further assessing the existence of interactions of hGal3
CRD with the carbohydrate analogue. Indeed, both positive
and negative chemical shift changes were observed for many
1H−15N signals of hGal-3 upon addition of 5 loaded with the
different metal ions. In addition, depending on the type of
saturation was received at protons belonging to the galactose
unit, especially protons 4−6. A very similar pattern is obtained
in STD experiments performed with lactose in the presence
of the same protein (Figure 4, C). Taken together, these data
indicate that the introduction of the PhDTA linker does not
alter the recognition of the sugar by its natural receptor and
that the interaction of the conjugate takes place at the level of
the sugar moiety.
8015
dx.doi.org/10.1021/ja502406x | J. Am. Chem. Soc. 2014, 136, 8011−8017

Journal of the American Chemical Society
Article
metal, some 1H−15N cross peaks became broadened and others
even disappeared from the spectra, due to PREs. This effect was
especially drastic in the case of Dy3+, as expected from its large
paramagnetic effects.19 The overlay of the control spectrum
acquired with La3+ with the corresponding spectrum acquired
with Tm3+ is shown in Figure 5.
observed again (see Supporting Information). This result shows
that 5 and lactose compete for the same binding site, further
assessing the existence of equivalent binding modes, and,
therefore, that the interaction of 5 with hGal-3 is specific.
Finally, MSpin calculations were carried out to further confirm
that the experimental PCSs for the hGal3 CRD in the presence of
5 are in agreement with those expected for the known X-ray
structure of the hGal3/lactose complex. The hGal3/lactose
complex PDB 2NN8 was used as starting point to build the
structure of hGal3/5 complex. In these coordinates, lactose is in
the synΦ/synΨ conformation. Different rotamers of the
biphenilic moiety and the bond connecting the amide adjacent
to the aromatic ring were considered to take into account the
linker flexibility in the calculations, as described above for the
lactose derivative in the free state. In all cases, the back-calculated
PCSs for the protein signals were in very good agreement with
the experimental data. MSpin Q factors were 0.3 and 0.2 for Tm3+
and Dy3+, respectively (Figure 5C and Supporting Information).
The results were analyzed on the basis of the available three-
dimensional structure of the complex using the X-ray coordinates
of the CRD of the hGal-3/lactose complex (PDB code 2NN8).47
This geometry was used as template to generate the putative
complex with compound 5, by attaching the LBT fragment to
the lactose moiety, as described in the Methods section.
Interestingly, PCS values higher than 0.02 can be observed for
32 NH signals, even in residues that are 42 Å apart from the
metal core (Supporting Information). This data set represents an
improvement with respect to the results described with nitroxide
spin-labeled carbohydrate analogues.20 In that case, only
perturbation of residues that fall within a radius of 20 Å from
the spin label could be detected, and in addition, only intensity
losses could be measured for obtaining structural information.
As we now report, the use of lanthanide-binding tags attached to
the carbohydrate could allow for the measurement of both PREs
and PCSs. This fact, in principle, could be used to detect very
weak interactions or lead to a more refined structural analysis.
The PCS data obtained for Dy3+ and Tm3+ exhibited a
different pattern of positive and negative values, as expected
from their corresponding iso-surface shapes. The obtained PCS
values are given in the Supporting Information.
The most affected residues upon addition of 5 loaded with
Tm3+ are I171, C173, E185, and Q187, (PCSs: −0.06, −0.07,
−0.06, and −0.06 ppm, respectively). These protons are located
at shorter distances respect to the metal (23.6, 20.2, 19.1,
21.9 Å, respectively). In the case of Dy3+, these signals disappear
due to the higher PRE effect of this ion. In this case, the largest
PCSs were detected for A146, L147, D148, F157, and S237
(0.04, 0.05, 0.05, 0.05 and 0.05, respectively). These protons are
located at 29.7, 27.8, 29.7, 25.9, and 31.3 Å, respectively.
A hallmark of ligand binding in galectins is the C−H/
π-interaction of the galactose unit with a properly located Trp
residue.2 This interaction provides specificity and is also useful
as sensor for fluorescence-based monitoring of binding.48 In the
case of h-Gal3, W181 and the galactose moiety of lactose are in
close contact. Due to its proximity to the metal ion (NH-metal
distance 16 Å), the backbone NH of this residue is one of the
most affected upon ligand binding and disappears in all the
spectra acquired with the carbohydrate conjugate loaded with
the paramagnetic metal ions (Tm3+or Dy3+). It is worth noting
that this is one of the few signals that becomes broadened below
detection in the experiment with Tm3+. In fact, the data measured
with the carbohydrate conjugate loaded with Tm3+ (metal with
the lowest PRE of the two employed) were very useful to
characterize the protein−carbohydrate interaction surface. There-
fore, the combination of data obtained with the different metals
led to an improved definition of the binding event.
CONCLUSION
■
A novel lanthanide-binding tag has been designed to function-
alize carbohydrate moieties maintaining their possible bio-
activity via lectin binding. Loading the chelating part with two
paramagnetic ions enabled to efficiently determine the
conformation of a model disaccharide (lactose) by using PCS
data, independent of NOEs and J-couplings. Beyond conforma-
tional characterization of the ligand, bound-state topological
features of a carbohydrate binding protein can also be defined.
Paramagnetic effects of the metal ion are sensed by the protein
as evidenced in the PCS values of the cross peak signals in
1H−15N HSQC spectra. In this case, both PRE and PCS effects
can be transferred from the LBT-decorated carbohydrate to the
protein, provided that the protein accommodates the carbohy-
drate. Since the chemical shift perturbations induced by the
carbohydrate at the binding site of the protein and its periphery
are enhanced by the PCS effects of the metal present at the LBT,
detection of molecular interactions can readily be performed. This
approach may be particularly useful for detecting low-affinity
events or transient interactions. Furthermore, a detailed analysis
of the data may permit the proper location of the binding site,
provided that a putative geometrical model is available.
Using this methodology, screening for lectin-blocking com-
pounds can thus be advanced, since detection of secondary sites
can be accomplished. Of particular relevance, comparative analysis
within a lectin family, in our case adhesion/growth-regulatory
galectins, is readily possible with such a sensor, probing into lectin
differences. Moreover, feasibility of studies on other carbohydrate-
binding proteins such as glycoenzymes or transporters markedly
broadens the scope of possible applications.
Finally, this approach constitutes an attractive alternative to
the use of protein tagging. In our case, no protein engineering is
required, thus allowing for the study of the protein in its native
form.
ASSOCIATED CONTENT
* Supporting Information
Finally, in order to rule out the possibility of a different
binding mode for 5 to that of lactose, competition experiments
with lactose were carried out. Lactose was added to hGal-3
samples containing 5. In the presence of lactose, the amide
■
S
Protein purification protocol. Synthesis and chemical character-
ization of 5 and intermediates. Dissociation constant
calculations obtained from the NMR titrations. Description of
the MSpin calculations. Detailed table with the PCSs measured
with the different metal ions (Tm3+ and Dy3+) and parameters
of the calculated Δχ tensors. This material is available free of
charge via the Internet at http://pubs.acs.org.
1
NH signals in the H−15N HSQC spectrum of hGal-3 shifted
back toward the chemical shift values of the blank sample
(lactose loaded with La3+). Furthermore, those cross peaks
that had disappeared or had shown a large line broadening in
the presence of 5 loaded with either Tm3+ or Dy3+ could be
8016
dx.doi.org/10.1021/ja502406x | J. Am. Chem. Soc. 2014, 136, 8011−8017

Journal of the American Chemical Society
Article
(19) Otting, G. J. Biomol. NMR 2008, 42, 1−9.
AUTHOR INFORMATION
■
(20) Jain, N. U.; Venot, A.; Umemoto, K.; Leffler, H.; Prestegard, J.
H. Protein Sci. 2001, 10, 2393−400.
Corresponding Author
jjbarbero@cib.csic.es
(21) Zhuang, T.; Leffler, H.; Prestegard, J. H. Protein Sci. 2006, 15,
1780−90.
Notes
(22) Seidel, R. D., 3rd; Zhuang, T.; Prestegard, J. H. J. Am. Chem. Soc.
2007, 129, 4834−9.
The authors declare no competing financial interest.
(23) Zhuang, T.; Lee, H. S.; Imperiali, B.; Prestegard, J. H. Protein Sci.
2008, 17, 1220−31.
ACKNOWLEDGMENTS
■
This research was financially supported by funding from the EC
Seventh framework program FP7/2007-2013 under REA grant
agreements no. 260600 (GlycoHIT) and 317297 (GLYCO-
PHARM), MHit project of the Comunidad de Madrid, Spain,
CTQ2012-32025, CTQ2012-31063 projects, and Ramon y Cajal
grant to Dr. Angeles Canales from the Spanish government. We
would also like to thank the NMR facility of the Universidad
Complutense de Madrid (CAI) and the company MestreLab
Research for making MSpin software available. M.A.B. acknowl-
edges an FPI fellowship from the Spanish Ministry of Economy
and Competitiveness.
(24) Amano, M.; Eriksson, H.; Manning, J. C.; Detjen, K. M.; Andre,
́
S.; Nishimura, S.; Lehtio, J.; Gabius, H. J. FEBS J. 2012, 279, 4062−80.
(25) Diez-Revuelta, N.; Velasco, S.; Andre, S.; Kaltner, H.; Kubler,
́
̈
D.; Gabius, H. J.; Abad-Rodriguez, J. J. Cell. Sci. 2010, 123, 671−81.
(26) Kaltner, H.; Gabius, H. J. Histol. Histopathol. 2012, 27, 397−
416.
́
(27) Krzeminski, M.; Singh, T.; Andre, S.; Lensch, M.; Wu, A. M.;
Bonvin, A. M.; Gabius, H. J. Biochim. Biophys. Acta 2011, 1810, 150−
61.
(28) Quattroni, P.; Li, Y.; Lucchesi, D.; Lucas, S.; Hood, D. W.;
Herrmann, M.; Gabius, H. J.; Tang, C. M.; Exley, R. M. Cell Microbiol.
2012, 14, 1657−75.
(29) Sanchez-Ruderisch, H.; Fischer, C.; Detjen, K. M.; Welzel, M.;
Wimmel, A.; Manning, J. C.; Andre,
277, 3552−63.
(30) Saio, T.; Ogura, K.; Shimizu, K.; Yokochi, M.; Burke, T. R., Jr.;
Inagaki, F. J. Biomol. NMR 2011, 51, 395−408.
́
S.; Gabius, H. J. FEBS J. 2010,
REFERENCES
■
(1) Gabius, H. J. The Sugar Code: Fundamentals of Glycosciences
Wiley-VCH: Weinheim, 2009.
(2) Gabius, H. J.; Andre, S.; Jimenez-Barbero, J.; Romero, A.; Solis,
́ ́
D. Trends Biochem. Sci. 2011, 36, 298−313.
(3) Chabre, Y. M.; Giguere, D.; Blanchard, B.; Rodrigue, J.;
Rocheleau, S.; Neault, M.; Rauthu, S.; Papadopoulos, A.; Arnold, A.
A.; Imberty, A.; Roy, R. Chemistry 2011, 17, 6545−62.
(4) Prost, L. R.; Grim, J. C.; Tonelli, M.; Kiessling, L. L. ACS Chem.
Biol. 2012, 7, 1603−8.
́
(5) Arda, A.; Blasco, P.; Varon Silva, D.; Schubert, V.; Andre, S.;
Bruix, M.; Canada, F. J.; Gabius, H. J.; Unverzagt, C.; Jimenez-Barbero,
J. J. Am. Chem. Soc. 2013, 135, 2667−75.
(6) Canales, A.; Lozano, R.; Lopez-Mendez, B.; Angulo, J.; Ojeda, R.;
Nieto, P. M.; Martin-Lomas, M.; Gimenez-Gallego, G.; Jimenez-
Barbero, J. FEBS J. 2006, 273, 4716−27.
(7) Garcia-Mayoral, M. F.; Canales, A.; Diaz, D.; Lopez-Prados, J.;
Moussaoui, M.; de Paz, J. L.; Angulo, J.; Nieto, P. M.; Jimenez-
Barbero, J.; Boix, E.; Bruix, M. ACS Chem. Biol. 2013, 8, 144−51.
(8) Neuhaus, D.; Williamson, M. P. The Nuclear Overhauser Effect in
Structural and Conformational Analysis; Wiley-VCH: New York, 2000.
(9) Prestegard, J. H.; Bougault, C. M.; Kishore, A. I. Chem. Rev. 2004,
104, 3519.
(31) Mallagaray, A.; Canales, A.; Dominguez, G.; Jimenez-Barbero, J.;
Perez-Castells, J. Chem. Commun. (Camb) 2011, 47, 7179−81.
(32) Erdelyi, M.; d’Auvergne, E.; Navarro-Vazquez, A.; Leonov, A.;
Griesinger, C. Chemistry 2011, 17, 9368−76.
(33) (a) Yamamoto, S.; Yamaguchi, T.; Erdelyi, M.; Griesinger, C.;
Kato, K. Chemistry 2011, 17, 9280−2. (b) Canales, A.; Mallagaray, A.;
Per
́
ez-Castells, J.; Boos, I.; Unverzagt, C.; Andre,
́
S.; Gabius, H. J.;
Canada, F. J.; Jimen
13789−93.
́
ez-Barbero, J. Angew. Chem., Int. Ed. 2013, 52,
̃
(34) (a) MSpin; Mestrelab Research: Santiago de Compostela, Spain;
http://www.mestrelab.com; accessed May 5, 2014. (b) Navarro-
Vázquez, A. Magn. Reson. Chem. 2012, 50, S73−S79.
(35) Asensio, J. L.; Jimenez-Barbero, J. Biopolymers 1995, 35, 55−73.
(36) Asensio, J. L.; Martin-Pastor, M.; Jimenez-Barbero, J. Int. J. Biol.
Macromol. 1995, 17, 137−48.
(37) Asensio, J. L.; Martin-Pastor, M.; Jimenez-Barbero, J. J. Mol.
Struct.: THEOCHEM 1997, 395−396, 245−270.
(38) Martín-Pastor, M.; Espinosa, J. F.; Asensio, J. L.; Jimenez-
Barbero, J. Carbohydr. Res. 1997, 298, 15−49.
(39) Martin-Pastor, M.; Canales, A.; Corzana, F.; Asensio, J. L.;
Jimenez-Barbero, J. J. Am. Chem. Soc. 2005, 127, 3589−95.
(40) Mizuno, M.; Funahashi, S.; Nakasuka, N.; Tanaka, M. Inorg.
Chem. 1991, 30, 1550−1553.
(41) Mayer, M.; Meyer, B. Angew. Chem., Int. Ed. 1999, 38, 1784−
1788.
(10) Xia, J.; Margulis, C. J.; Case, D. A. J. Am. Chem. Soc. 2011, 133,
15252.
(11) Roldos, V.; Canada, F. J.; Jimenez-Barbero, J. ChemBioChem
2011, 12, 990.
(12) Bernardi, A.; Jimenez-Barbero, J.; Casnati, A.; De Castro, C.;
Darbre, T.; Fieschi, F.; Finne, J.; Funken, H.; Jaeger, K. E.; Lahmann,
M.; Lindhorst, T. K.; Marradi, M.; Messner, P.; Molinaro, A.; Murphy,
P. V.; Nativi, C.; Oscarson, S.; Penades, S.; Peri, F.; Pieters, R. J.;
Renaudet, O.; Reymond, J. L.; Richichi, B.; Rojo, J.; Sansone, F.;
Schaffer, C.; Turnbull, W. B.; Velasco-Torrijos, T.; Vidal, S.; Vincent,
S.; Wennekes, T.; Zuilhof, H.; Imberty, A. Chem. Soc. Rev. 2013, 42,
4709.
(42) Zhang, Y.; Yamamoto, S.; Yamaguchi, T.; Kato, K. Molecules
2012, 17, 6658−71.
(43) Tvaroska, I.; Bleha, T. Adv. Carbohydr. Chem. Biochem. 1989, 47,
45−123.
(44) Losonczi, J. A.; Andrec, M.; Fischer, M. W. F.; Prestegard, J. H.
J. Magn. Reson. 1999, 138, 334−342.
(45) Meyer, B.; Peters, T. Angew. Chem., Int. Ed. Engl. 2003, 42, 864−
(13) Gabius, H. J.; Siebert, H. C.; Andre,
́
S.; Jimenez-Barbero, J.;
90.
Rudiger, H. Chembiochem 2004, 5, 740−64.
̈
(46) Shuker, S. B.; Hajduk, P. J.; Meadows, R. P.; Fesik, S. W. Science
1996, 274, 1531−4.
́
(14) Jimenez Barbero, J.; Peters, T. NMR Spectroscopy of
Glycoconjugates; Wiley-VCH, Weinheim, 2002.
(47) Collins, P. M.; Hidari, K. I.; Blanchard, H. Acta Crystallogr., Sect.
D: Biol. Crystallogr. 2007, 63, 415−9.
(15) Koenig, S. H.; Brown, R. D., 3rd; Brewer, C. F. Proc. Natl. Acad.
Sci. U S A 1973, 70, 475−9.
(48) Asensio, J. L.; Arda, A.; Canada, F. J.; Jimenez-Barbero, J. Acc.
Chem. Res. 2013, 46, 946−54.
(16) Barbieri, R.; Bertini, I.; Cavallaro, G.; Lee, Y. M.; Luchinat, C.;
Rosato, A. J. Am. Chem. Soc. 2002, 124, 5581−7.
(17) Leonov, A.; Voigt, B.; Rodriguez-Castaneda, F.; Sakhaii, P.;
Griesinger, C. Chemistry 2005, 11, 3342−8.
(18) Saio, T.; Ogura, K.; Yokochi, M.; Kobashigawa, Y.; Inagaki, F. J.
Biomol. NMR 2009, 44, 157−66.
8017
dx.doi.org/10.1021/ja502406x | J. Am. Chem. Soc. 2014, 136, 8011−8017