NMR study of a Lewis(X) pentasaccharide derivative: Solution structure and interaction with cations

Carbohydrate Research 315 (1999) 48–62

NMR study of a Lewis^Xpentasaccharide derivative:

solution structure and interaction with cations

Benoˆıt Henry ^a, Herve´ Desvaux ^a, Marina Pristchepa ^a, Patrick Berthault ^a,*,

b

Yong-Min Zhang ^b, Jean-Maurice Mallet ^b, Jacques Esnault ^b, Pierre Sinay¨

^aCEA Saclay, Laboratoire Commun de RMN, DSM/DRECAM/SCM, F-91191 Gif sur Y6ette, France

^bEcole Normale Supe´rieure, De´partement de Chimie, Associe´ au CNRS, 24 rue Lhomond, F-75231 Paris, France

Received 7 October 1998; accepted 19 November 1998

Abstract

The structure and conformation of the synthetic pentasaccharide Gal(b1-4){Fuc(a1-3)}GlcNAc(b1-3)Gal(b1-

4)Glc-bOMe of the Lewis^Xfamily has been determined by NMR spectroscopy in dimethyl sulfoxide and methanol.

In these solvents, the binding constants with calcium have been evaluated as 9.5 and 29.6 M⁻¹, respectively. Study

of the interaction sites has been achieved through the use of paramagnetic divalent cations and distance triangulation

methods. Two regions have been found, the ﬁrst one in the vicinity of the fucose unit, the second one closer to the

1

Keywords: Lewis^X; Oligosaccharide; Paramagnetic ions; Calcium; H NMR; ¹³C NMR

typic interaction between two neutral

oligosaccharides is mediated by the presence

of Ca²⁺, Mg²⁺, or Mn²⁺. The association of

neutral carbohydrates and polyols with metal

cations in solution has been studied [6–8],

using a wide range of analytical techniques:

electrophoresis [9,10], thin-layer chromatogra-

phy [11], Fourier-transform infrared spec-

1. Introduction

The molecular basis of cell recognition and

adhesion is of paramount importance for un-

derstanding biological phenomena such as dif-

ferentiation, organisation and development in

multicellular organisms, as well as diseases

involving abnormal cell adhesion. It has been

demonstrated that speciﬁc interaction between

carbohydrates may be an important initial

step in such cellular processes [1–3].

troscopy

[12],

microcalorimetry

[13],

ebulliometry [6], luminescence excitation spec-

troscopy [14], pHmetry [15], etc., but it is only

when nuclear magnetic resonance spec-

troscopy [16] or mass spectrometry [17–19]

has been used that more precise information

about the structure of the complex could be

obtained.

Although high-resolution liquid-state NMR

spectroscopy is a very powerful tool for eluci-

dation of intermolecular events, evidencing

carbohydrate–carbohydrate interaction such

as Le^x–Le^xwith the help of this spectroscopy

Speciﬁc interaction between Le^xand Le^x

determinants of glycosphingolipids has been

studied by Hakomori et al. [4,5]. It may

provide an interesting possible basis for cell

recognition in preimplantation embryos and

in embryonal carcinoma cells. Such a homo-

* Corresponding author. Tel.: +33-1-69084245; fax: +33-

1-69089806.

E-mail address: patrick.berthault@cea.fr (P. Berthault)

PII: S0008-6215(98)00301-2

B. Henry et al. / Carbohydrate Research 315 (1999) 48–62

49

Scheme 1.

remains difﬁcult. Indeed this phenomenon

takes proﬁt in situ from a cooperative effect,

made possible by the high density of sugar

headgroups at the cell surface. Some attempts

to mimic this glycoside-enriched surface for

NMR studies have made use of ganglioside

micelles. However, even in such conditions, it

has not been possible to evidence carbohy-

drate–carbohydrate interactions [20].

The ﬁrst experiments deal with the precise

analysis of the structure and internal dynamics

of this oligosaccharide in solution through

recent dedicated approaches [23–25]. Inﬂu-

ence of Ca²⁺ions on the proton spectra is

then studied in order to derive the correspond-

ing association constants. A third type of ex-

periment uses paramagnetic ions to mimic

complexation of calcium ions as detected by

¹H and ¹³C NMR, and to allow the determi-

nation of their location relative to structure 1.

It appears to us that a knowledge of the

structure, internal dynamics in solution, and

interaction with divalent cations of Le^X

derivatives through NMR is a prerequisite to

the study of site-speciﬁc homotypic interac-

tion. Wormald et al. have failed in showing

speciﬁc or general interaction of Le^Xand the

pentasaccharide (Gal(b1-4){Fuc(a1-3)}GlcN-

Ac(b1-3)Gal(b1-4)Glc) oligosaccharides with

1

Ca²⁺or Mn²⁺[21]. Their study used H

NMR in D₂O, and no detectable change in the

proton chemical shifts of the sugars was ob-

served. As hydroxyl groups represent the ex-

ternal surface of an oligosaccharide, the

signals of their protons are expected to experi-

ence more signiﬁcant effects upon weak inter-

action with an ion than the non-labile ones.

For this reason, we have undertaken an NMR

study of a pentasaccharide containing the Le^X

motif, in dimethyl sulfoxide and in methanol,

where these signals can easily be observed.

Moreover, the binding constants with calcium

are expected to be greater in methanol than in

water [22].

2. Experimental

Synthesis of pentasaccharide 1.—Com-

pound 1 was prepared efﬁciently as follows:

reaction in CH₂Cl₂of bromide 2 [26] (Scheme

1) with the known diol 3 [27], using silver

triﬂuoromethanesulfonate (AgOTf) as the pro-

moter, gave the disaccharide 4.

The regiochemistry of the newly introduced

glycosidic linkage was conﬁrmed by the H

1

NMR data which show a signal for Glc H-3

at 4.64 ppm (ddd, J_3,OH1.2 Hz, J_3,48.6 Hz, J_2,3

10.4 Hz). Its stereochemistry was determined

to be b on the basis of the Gal H-1–H-2

coupling constant (J_1,28.1 Hz). The fucosyla-

tion of 4 with ethyl 2,3,4-tri-O-benzyl-1-thio-

The molecular structure of this pentasac-

charide 1 is very close to the Wormald et al.

pentasaccharide, except that the terminal glu-

cose unit retains the b-conﬁguration as methyl

glycoside.

b- -fucopyranoside 5 [28] in the presence of

L

50

B. Henry et al. / Carbohydrate Research 315 (1999) 48–62

Scheme 2.

N-iodosuccinimide (NIS)–triﬂuoromethane-

sulfonic acid (TfOH) [29] provided the trisac-

charide 6 in 86% yield. It was therefore

possible to activate the armed SEt group with-

out affecting the disarmed SPh group by using

1 equiv of NIS [30]. The stereochemistry of

the newly introduced glycosidic linkage was

determined to be a on the basis of the low

value of the Fuc H-1–H-2 coupling constant

(J_1,23.5 Hz).

Tesla. Assignment of the proton chemical

shifts of 1 was performed through homonu-

clear and heteronuclear 1D and 2D NMR

experiments in CD₃OH and CD₃OD. The

1

main H NMR experiments were COSY, RE-

LAYs and TOCSY for determination of the

intra-unit through-bond connectivities, and

off-resonance ROESY for the inter-unit se-

quencing. Carbon assignment was achieved by

inverse HMQC [32,33] and HMBC [34]

experiments.

The coupling of 6 and methyl 2,3,6-tri-O-

benzyl-4-O-(2,6-di-O-benzyl-b-

D

-galactopyran-

The conformation and the internal dynam-

ics of the oligosaccharide in methanol were

determined by the use of a proton NMR

method that was recently proposed by us [25].

It has the advantage of avoiding the require-

ment of an internal distance reference to

derive the other interproton distances. For

this purpose, 27 2D off-resonance ROESY

experiments with effective spin-lock ﬂip angles

ranging from 0 to 54.7° and mixing times

between 50 and 150 ms were performed, as

well as numerous 1D off-resonance ROESY

to complete them when a high resolution was

needed. After assumption of a simple mo-

tional model, where the dipolar auto-correla-

tion function associated to each proton pair

can be described by an exponential function,

66 H–H distance constraints and 66 local

correlation times were derived for the pen-

tasaccharide in the absence of calcium. For

pentasaccharide 1 in the presence of calcium,

as most of the hydroxyl proton signals were

broadened by exchange with the solvent, they

do not give rise to dipole–dipole cross-peaks

with the neighbouring protons. Therefore only

49 distance constraints are derived in this

case.

osyl)-b-

D

-glucopyranoside 7, easily prepared

-lactoside [31], was performed

from methyl b-

D

in the conditions described above, generating

the pentasaccharide 8 in 86% yield (Scheme 2).

The stereochemistry of the newly intro-

duced linkage was determined to be b on the

basis of the GlcN H-1–H-2 coupling constant

(J_1,28.5 Hz). The regioselectivity of the glyco-

sylation was conﬁrmed by acetylation of 8.

1

The H NMR spectrum of the acetate showed

the presence of H-4 of the galactose unit at

5.42 ppm (dd, J_4,5B1 Hz, J_3,4=3.5 Hz), indi-

cating the position of the new glycosidic link-

age in 8 to be at OH-3 of the acceptor 7.

Classical deprotections gave the expected pen-

tasaccharide 1. A detailed experimental part

will be published elsewhere.

NMR spectroscopy.—Before each NMR ex-

periment, the samples were freeze-dried and

accurately weighed in the NMR tube. They

were then dissolved either in (CD₃)₂SO, in

CD₃OH, or in CD₃OD (all originating from

CEA, Eurisotop). All experiments were per-

formed on Bruker DRX spectrometers

1

equipped with self-shielded gradient H/¹³C/

¹⁵N 5-mm inverse probeheads, at 11.7 and 14

B. Henry et al. / Carbohydrate Research 315 (1999) 48–62

51

Molecular modelling.—The H–H distance

Dl_obs=l_obs−l_free

=(l_complex−l_free) K[Ca] (1+K[Ca])⁻¹

(2)

constraints obtained after off-resonance

ROESY are used in a simulated annealing

procedure using X-PLOR 3.1 software facilities

[35]. Starting 50 times from a different ran-

dom conformation, 50 structures were gener-

ated after 35 ps of high temperature (1000 K)

dynamics and slow cooling during 28 ps, both

with a timestep of 3.5 fs. During the high-

temperature dynamics, progressive introduc-

tion of the distance constraints with low

weight on the van der Waals terms ensured a

large sampling of the conformational space.

The slow cooling was designed to restore the

correct values of the covalent energy terms.

According to criteria based on the NMR con-

straints and covalent energy terms, some re-

sulting structures were rejected, and the

remaining set was used to compute the aver-

age structure and the rms deviations on the

coordinates.

An iterative program based on the Marquardt

algorithm [36] allows the determination of the

association constant K as well as l_complex, the

chemical shift of each proton in the pure

complex. The ﬁt used simultaneously every

peak exhibiting a chemical shift variation. In

this way a large oversampling of the system

was secured.

Use of paramagnetic ions in NMR.—To

avoid any inner-sphere complexation with the

cations studied, chloride was chosen as a

counter-ion for every experiment [37,38].

13

Mn²⁺. Four C longitudinal self-relaxation

experiments were performed with equivalents

of MnCl₂ranging from 0 to 0.025 (0, 0.005,

0.015, 0.025). The initial concentration of 1 in

methanol was 57 mM. The dilution effect

upon addition of MnCl₂·2H₂O was negligible.

Co²⁺. A 2D COSY experiment was used to

univocally assign the maximum of proton sig-

nals in the presence of 0.5 equiv of CoCl₂.

Due to line broadening at higher concentra-

tions, we were not able to determine the equi-

librium constant with the cation, and only

one addition of cobalt was used for deter-

mining the interaction sites. The concentration

of 1 in methanol was 9.2 mM. The dilution

effect upon addition of CoCl₂·2H₂O was negli-

gible.

Calcium titration.—In (CD₃)₂SO, the start-

ing solution was constituted of 1 (3.51 mg) in

400 mL at 295 K (concentration: 10.1 mM).

From 0 to 37.9 equiv of CaCl₂·2H₂O were

added at two concentrations: 808 mM and

80.8 mM. In CD₃OD, the starting solution

was constituted of 1 (3.91 mg) in 360 mL at

290 K (concentration: 12.5 mM). From 0 to

7.55 equiv of CaCl₂·2H₂O were added from a

solution of 120 mM.

Correcting the dilution and pH variation

effects, the observed chemical shift variation

of a given signal upon addition of an ion in

solution can be expressed as:

Location of Mn²⁺and Co²⁺.—On each of

the accepted structures after simulated anneal-

ing, the Cartesian coordinates of two Mn²⁺

ions were randomly drawn in order to ensure

coverage of the subspace of solutions as com-

plete as possible, and to allow a statistical

analysis. The most probable location of the

ions was determined through a simplex proce-

dure [36] which minimised the following target

function, using the relative variations of the

¹³C self-relaxation rates D(1/T₁ⁱ) due to addi-

Dl_obs=Dl_struc+Dl_dia+Dl_{pseudo-contact}

+Dl_contact

(1)

where Dl_strucrepresents the chemical shift

variation due to conformational modiﬁcation

of the molecule upon addition of the ion, Dl_dia

the diamagnetic susceptibility shift, and

Dl_{pseudo-contact}

,

Dl_contactthe paramagnetic

tion of Mn²⁺

:

pseudo-contact and contact hyperﬁne cou-

pling terms, respectively. In the case of cal-

cium, the last two terms vanish. As fast

exchange conditions are fulﬁlled and assuming

a 1:1 complex, Dl_obscan also be expressed as,

after correction of the global susceptibility

shift of the sample:

Á

Â²

ꢀ₁ꢁ

D

Ã

₋₆Ã

r⁻_I⁶+r⁻⁶

ref

II

1

T₁ⁱ

r⁻_I⁶+r_II

i

²=% Ã

−

Ã

(3)

2

ꢀ ₁ꢁ

|

ref

i

Ã

D

T^r₁^ef

Ä

Å

52

B. Henry et al. / Carbohydrate Research 315 (1999) 48–62

The summation was done on all the carbons

for which the longitudinal self-relaxation

rates had been measured (all except those

exhibiting overlapped signals). The back cal-

culated distance between the Mn²⁺ion and

the ith carbon (subscripts I and II refer to

the manganese ion considered) is denoted as

rⁱ. The reference (subscript ref) was taken as

the carbon for which the largest variation on

(1/T₁) was observed (i.e. Gal^IC-6). |_iwas

the estimated uncertainty. The consideration

of different populations for the two locations

of the cation does not induce a signiﬁcant

axial symmetry [39]. Due to the too high

number of parameters to adjust in a direct

ﬁt, it was difﬁcult to be sure that the ran-

dom drawing of the parameters was sufﬁ-

ciently wide to explore the complete

subspace of solutions, and then to perform

conﬁdent statistical tests. We consequently

used the following procedure. The results on

the best-ﬁt manganese coordinates were used

as starting points for the Co²⁺ions, and

only q_I, _I, q_II, were reﬁned in a ﬁrst

II

step. Then noise was added to the ten

parameters (the best-ﬁt coordinates found

for Mn²⁺and these four angles), and this

full set was again reﬁned by the same sim-

plex procedure against Eq. (4). As well as

for Mn²⁺, the introduction of a difference

in population between the two cobalt loca-

tions, which would mean a different bind-

ing constant or paramagnetic contribution,

does not increase signiﬁcantly the ², be-

cause of the poor ratio between the number

of experimental data and the number of

parameters to be minimised (42 vs. 10). The

solutions for the cobalt locations were then

treated by the same procedure as for

manganese.

Obviously, the location of the ion was

strongly dependent of the pentasaccharide

3D structure. For describing the results, two

types of distance average were measured. Ei-

ther for each 1 structure, the average loca-

tion of the ion, denoted X, was

considered to evaluate some representative

distances with atoms of the pentasaccharide.

Or each location of the ion obtained after

the minimisation step was used to calculate

the distances with atoms of the pentasaccha-

ride; then the complete ensemble average

was computed on all the locations and all

the structures possible for 1. These two

methods lead to results that are similar but

not identical, since the distributions are not

Gaussian.

2

improvement of the minimised value. The

resulting coordinates inside a cube 40×40×

3

˚

40 A centred on the molecular coordinates,

2

and with a low value, were kept and clus-

3

˚

tered in pieces of 0.2×0.2×0.2 A . Then

the average locations of the two ions were

computed, as well as the rmsd along the

coordinates.

For cobalt, the same principle as for man-

ganese was used. But instead of six variables

which correspond to the six coordinates (x_I,

y_I, z_I, x_II, y_II, z_II), there were ten variables to

consider (the same, plus q_I, _I, q_II, _II,

which deﬁne the principal axes of the para-

magnetic susceptibility tensors for the cobalt

ions in the pentasaccharide molecular

frame). The function to be minimised, which

takes into account the ratio of the paramag-

netic dipolar shifts, is:

3 cos²c _i−1 3 cos²c −1

Á

Â²

i

II

I

+

Ã

1

Dl_i

r_I³

r³_II

i

²=% ₂Ã

−

Ã

|

Dl_ref3 cos²c_I_ref−1 3 cos²c_II_ref−1

i

Ã

+

r_I³

r³_II

Ä

Å

ref

(4)

Dl_i, Dl_refare the chemical shift variations of

protons i and reference, respectively. The an-

gle between the principal axis of the Co²⁺

anisotropic susceptibility tensor and the vec-

tor r_ijoining the Co²⁺ion and the consid-

ered nucleus is denoted c. Similarly to what

was done for manganese, and in order to

minimise the relative error, the proton cho-

sen as reference was the one for which the

largest chemical shift variation was observed

(Gal^IH-6). In this operation, assumptions

were made that the contact term was negligi-

ble, and secondly that the cobalt ion was of

3. Results and discussion

Solution structure of pentasaccharide 1.—

The preliminary step before determination of

the 3D structure by NMR is the assignment

B. Henry et al. / Carbohydrate Research 315 (1999) 48–62

53

Table 1

¹³C and ¹H chemical shifts for 1 in (CD₃)₂SO at 290 K using (CD₃)₂SO as internal reference (¹³C, 39.5 ppm; ¹H, 2.50 ppm)^a

C-1

C-2

C-3

C-4

C-5

C-6

CH₃

Gal^I

101.93

98.34

101.15

103.26

103.35

70.45

64.08

55.46

69.07

72.86

73.21

69.17

75.14

82.09

74.55

67.08

71.40

73.32

66.86

74.55

74.39

46.54

75.43

74.81

80.27

59.55

Fuc^II

16.18

22.84

GlcNAc^III

Gal^IV

Glc^V

59.27

60.06

60.11

55.89

CH₃

H-1

H-2

H-3

H-4

H-5

H-6

H-6%

Gal^I

4.268

4.824

4.688

4.252

4.096

3.272

3.391

3.613

3.423

2.997

3.221

3.540

3.689

3.421

3.297

3.643

3.450

3.621

3.861

3.300

3.218

4.650

3.308

3.480

3.296

3.520

3.397

Fuc^II

0.985

1.805

GlcNAc^III

Gal^IV

Glc^V

3.791

3.524

3.753

3.660

3.480

3.605

3.390

HO-1

HO-2

HO-3

HO-4

HO-5

HO-6

4.485

Gal^I

5.010

3.082

4.737

4.247

4.240

4.107

Fuc^II

GlcNAc^III

Gal^IV

Glc^V

4.701

4.721

4.615

4.919

5.207

4.688

4.616

^aThe notations H-6, H-6% do not refer to pro-R and pro-S protons, as no stereospeciﬁc assignment has been done.

have potent NMR methods to get informa-

tion on the local dynamics.

of the proton and carbon chemical shifts.

Although this has been achieved both in

dimethyl sulfoxide and in methanol, in the

following development, for the conforma-

tional analysis of the oligosaccharide, em-

phasis has been put on the analysis in

methanol.

Comparison of the proton chemical shifts

of 1 in (CD₃)₂SO and in CD₃OD (Tables 1

and 2) shows no big difference between

them, which could indicate that the structure

is not strongly modiﬁed between the two sol-

vents.

As the common problem encountered in

the resolution of molecular structures

through high-resolution NMR is how to dif-

ferentiate a high ﬂexibility from a lack of

NMR data for a part of the molecule, we

preferred to avoid a direct comparison be-

tween the 3D structure of 1 in the presence

and in the absence of calcium in CD₃OH.

Indeed as the exchange rates with solvent

are rather different in both cases (see below),

it could lead to the erroneous interpretation

that the structure in the absence of calcium

is more rigid than the other one, even if we

A second argument for avoiding such a

comparison is that the accuracy on the inter-

proton distance information is not high

enough to draw conclusions on slight varia-

tions of the orientation of a unit relative to

another one. This comparison could be

made only if a drastic conformation change

occurred after addition of calcium, effect

which does not appear. Therefore, the fol-

lowing part of the text only presents the

molecular modelling results for the study of

the pentasaccharide alone or in the presence

of 11.4 equiv of calcium in CD₃OD; only

some indicative comparisons are given.

For the study of the pentasaccharide alone

in methanol, after the simulated annealing

procedure, only 27 structures were kept. The

remaining ones were rejected, because their

potential energy was too high. The rms devi-

ation for the coordinates of the heavy atoms

˚

of the accepted structures was 1.02 A.

For the study of the pentasaccharide 1 in

methanol in the presence of calcium, 35

structures were kept according to the same

54

B. Henry et al. / Carbohydrate Research 315 (1999) 48–62

Table 2

1

¹³C and H chemical shifts for 1 in CD₃OD at 290 K^a

C-1

C-2

C-3

C-4

C-5

C-6

CH₃

1

Gal^I

103.79

100.25

103.76

104.85

105.18

72.64

74.74

69.90

76.80

62.70

Fuc^II

GlcNAc^III

Gal^IV

Glc^V

69.82

57.73

71.51

74.57

71.09

71.05

83.80

76.25

73.64

76.27

69.79

80.28

67.53

76.49

78.48

76.44

16.68

23.19

61.02

62.55

61.56

57.29

H-1

H-2

H-3

H-4

H-5

H-6

H-6%

CH₃

1

Gal^I

4.478

5.098

4.733

4.408

4.244

3.550

3.683

3.975

3.645

3.264

3.500

3.912

3.903

3.596

3.561

3.839

3.764

3.951

4.101

3.614

3.491

4.890

3.468

3.635

3.438

3.712

3.806

Fuc^II

GlcNAc^III

Gal^IV

Glc^V

1.219

2.025

3.921

3.726

3.886

4.010

3.823

3.946

3.566

1+Ca

Gal^I

4.512

5.114

4.750

4.443

4.293

3.563

3.746

4.017

3.629

3.299

3.548

3.990

3.894

3.659

3.604

3.890

3.830

3.952

4.145

3.659

3.532

4.915

3.503

3.686

3.472

3.783

3.788

Fuc^II

1.233

2.065

GlcNAc^III

Gal^IV

Glc^V

3.917

3.778

3.898

4.049

3.817

3.952

3.574

De6iation

Gal^I

0.003

−0.015

−0.014

0.004

−0.018

0.032

0.011

−0.047

0.004

0.017

0.047

−0.040

0.032

0.020

0.035

−0.030

0.013

0.014

0.010

−0.006

0.004

0.020

0.003

0.042

−0.047

Fuc^II

−0.017

0.009

GlcNAc^III

Gal^IV

−0.035

0.021

−0.019

0.008

−0.037

−0.025

Glc^V

0.018

0.012

−0.023

^a1+Ca refers to the pentasaccharide 1 with 11.4 equiv of CaCl₂. As reference, the carbon and proton signals of the CH₃group

of methanol have been calibrated to (49.0; 3.310) ppm. The bottom part of the Table displays the relative deviations

Dl−Dl_average. When they are greater than the threshold 0.03, the corresponding text is bold.

criterion on the potential energy. The rms

deviation for the coordinates of their heavy

the local correlation times along the molecule

was found. It has to be mentioned, however,

that the off-resonance ROESY experiment

performed in speciﬁc conditions conﬁrms the

presence of a strong relaxation leakage in this

part of the molecule (measured on proton

GlcNAc H-1 and Gal^IVH-4). Indeed, when

the effective spin-lock ﬁeld makes an angle of

54°7% with the static ﬁeld, a signiﬁcant dis-

crepancy between the calculated and expected

ratios of cross-relaxation rate over self-relax-

ation rate Y was observed [23,41]. The differ-

ence between the value of 0.5 expected for a

purely dipolar relaxation and the experimental

value was more important in the presence of

calcium. In this last case, about 50% of the

self-relaxation of the anomeric proton of the

GlcNAc unit does not arise from dipolar re-

˚

atoms was 0.94 A.

In both cases, only one or two nOe viola-

tions (over 48 and 49 distance constraints,

respectively) were found. These discrepancies

concern the distances between the anomeric

proton of GlcNAc and protons H-2 and H-4

of Gal^IV. This corresponds to the cleavage

between the rigid Le^xmotif and the lactose

part of the molecule. Therefore, the co-exis-

tence of two conformational families in solu-

tion could explain the impossibility of

representing the experimental data with a sim-

ple one-population model. We cannot how-

ever prove this point with the NMR data, as,

conversely to the case of methyl b-4-carbalac-

toside in water [40], no signiﬁcant variation in

B. Henry et al. / Carbohydrate Research 315 (1999) 48–62

55

laxation, while for all studied protons other

than GlcNAc H-1 and Gal H-4, the ratio Y

was close to 0.5. An attempt to ascribe this

discrepancy to fast conformational exchange

was not conclusive. The reason for this

could be that the exchange rate is higher

than about 3 MHz, which falls outside the

area spanned by our approach.

larly good afﬁnity with calcium [9]. These

behaviours led us to undertake the study of

calcium complexation in methanol and

dimethyl sulfoxide.

A simple comparison between the proton

spectra of 1 in methanol in the absence and

in the presence of calcium (Fig. 1) leads to

the following remarks: (i) as expected with

the diamagnetic feature of Ca²⁺, the chemi-

cal shift variations are small, but they exist

and can be used for determination of the

association constant (see below); (ii) no addi-

tional lines are observed after addition of

the metal, showing that the complexation is

fast on the NMR time scale, which is in

concordance with literature (the unique case

of slow kinetic reported so far is the associa-

tion of muellitol with lanthanides(III)) [8];

(iii) the exchange rates of the hydroxyl pro-

tons with solvent are totally modiﬁed, as al-

ready noticed by Symons et al. [42,43]. We

have checked that it is not due to a pH

effect.

Calcium titration.—In aqueous media,

neutral carbohydrates and polyols form

weak complexes with metal cations (K=1–

10 M⁻¹). However, in alcoholic media, the

association was usually stronger than in wa-

ter (up to 100-times higher for some fura-

nosides). The cation also inﬂuences the

reaction by its charge (stronger association

with M(III)) and its size (ideal ionic radius

+

2+

3+

˚

seems to be around 1 A (Na , Ca , La )

[8]. Polyols and sugars then have a particu-

To obtain more quantitative values, these

exchange rates have been determined using

dedicated 1D proton NMR experiments

(Fig. 2). In the absence of calcium, the ex-

change rates of the hydroxyl protons with

the solvent were measured around 10 Hz

(between 7.6 and 14 Hz). In the presence of

calcium, the rates are multiplied by at least

a factor 10, except for a particular hydroxyl

proton: HO-2 of the Gal^IVunit, which

passes from 10 to 35 Hz. This proton seems

to be involved in a hydrogen bond with the

oxygen of the carbonyl group of the Glc-

NAc unit.

As weak binding constants are expected,

care should be taken about the different ef-

fects acting on the carbohydrate spectrum

during addition of calcium. Beyond the com-

plexation effect, other phenomena such as

dilution effect, pH change and variation of

macroscopic magnetisation susceptibility can

interfere on the chemical shifts. A careful

systematic analysis was thus undertaken

prior to NMR calcium titration experiments.

The dilution effect on the proton chemical

shifts was measured independently. This was

necessary, as, each calcium chloride molecule

being accompanied by two water molecules,

Fig. 1. 500 MHz proton spectrum of 1 at 290 K in CD₃OH

(a), and in the presence of 11.4 equiv of CaCl₂(b). Both

spectra are obtained with the 3-9-19 WATERGATE pulse

scheme [48,49]. The amide proton resonance as well as the

methyl signals are voluntarily not plotted.

56

B. Henry et al. / Carbohydrate Research 315 (1999) 48–62

nal becomes a triplet, which could lead to the

misleading interpretation that another peak

appears at the same frequency. The real rea-

son for this effect was that at this cation

concentration, the H-2 and H-3 protons res-

onate at the same frequency, and consequently

create very strong coupling conditions, modi-

fying the H-1 and H-4 signals.

The association constant values found for

calcium binding are 9.5 M⁻¹in (CD₃)₂SO

(using 30 proton resonances), and 29.6 M⁻¹

in CD₃OD (using six proton resonances). As-

suming an uncertainty of 1.5 Hz on the chem-

ical shift measurement, in the case of

2

methanol, the reduced value was 0.54 (59

degrees of freedom). In the case of (CD₃)₂SO

it was equal to 4.6 (637 degrees of freedom)

and reduces to 2.07 (401 degrees of freedom)

when only the experiments with less than 8

equiv of CaCl₂are kept. For these last two

ﬁts, the best-ﬁt values of the equilibrium con-

stant K are almost identical. In the case of

dimethyl sulfoxide, when a large amount of

calcium was added, the discrepancy revealed

Fig. 2. Measurement of the exchange rates of some hydroxyl

protons with methanol. (A) Pulse sequence based on the

WEXII ﬁlter scheme [50]. Wide and narrow ﬁlled boxes

represent hard 180 and 90° pulses, respectively. The hydroxyl

proton of methanol is selectively excited by a 10 ms Gaussian

pulse followed by the WEXII ﬁlter. Then, it experiences an

off-resonance irradiation with an effective spin-lock ﬁeld an-

gle of 35°18%, in order to reduce cross-relaxation effects and

keep only the exchange peaks [51]. The last step of the

sequence is the WATERGATE ﬁlter [48,49] designed to sup-

press the solvent peak. The value of the gradients are: G1=3,

G2=10, and G3=7 G cm⁻¹. (B) Example of a 1D subspec-

trum obtained with 100 ms mixing time.

2

by the reduced shows that either the exper-

imental biases was underestimated or the sim-

ple model of complexation between one Ca²⁺

ion and one molecule of 1 was not perfect.

This last assumption was corroborated by the

results of the study of the paramagnetic ions

for which the ﬁts of the experimental data

require two locations for the ion (see below).

it could be responsible for a slight chemical

shift variation even in dimethyl sulfoxide.

Moreover, to minimise it, various concentra-

tions of calcium solutions were used for differ-

ent parts of the titration curve. Also, care was

taken that the minor pH changes occurring

upon addition of the calcium solution did not

lead to signiﬁcant chemical shift variations of

the proton resonances. The variation of the

sample diamagnetic susceptibility led to a

global shift of the spectrum, and was conse-

quently taken into account.

Some chemical shift variations as a function

of the calcium equivalents used to derive the

association constant in dimethyl sulfoxide are

shown in Fig. 3. In both solvents, the chemical

shifts of some protons vary in the opposite

direction from the majority. As an anecdote, a

striking effect was observed on the spectrum

of 1 in methanol in the presence of 7 equiv of

calcium. The doublet of the GlcNAc H-1 sig-

Fig. 3. Chemical shift variation for some protons of 1 in

(CD₃)₂SO upon addition of CaCl₂. ", amide proton of

GlcNAc; +, OH-2 of Gal^IV; ꢀ, H-6 of Fuc; ×, OH-4 of

Fuc; ꢁ, OH-6 of Gal^I.

B. Henry et al. / Carbohydrate Research 315 (1999) 48–62

57

in the presence of calcium did not reveal the

presence of this ax, eq, ax arrangement.

For such a large excess of calcium, it be-

comes possible that 1:2 complexes with two

calcium ions are formed. The monotonic be-

haviour of the titration curves (Fig. 3), and

the uncertainty on the determination of the

experimental error, imply that we are not

conﬁdent in testing more complicated equi-

From the proton chemical shifts between

the pentasaccharide alone in methanol and

in the presence of calcium (Table 2), it could

be observed that only some H-2, H-3 and

few H-4, H-6 signals experience strong

chemical shift variations. According to the

chosen threshold, it is remarkable that the

H-2, H-3 and H-4 protons of the Gal^Iunit

did not experience such shifts.

2

librium models on the sole basis of the

value.

The measured binding constants K are

greater than those given by Angyal for

monosaccharides in water, but lower than

the values found for some furanosides in

methanol [8]. It is worth noting that oligo-

and polysaccharides are known to give

stronger complexes than the monosaccha-

rides from which they are issued [10], proba-

bly because of the increasing number of

binding sites available as well as the capa-

bility of cross-linking. In the present case,

the low K value indicates that the associa-

tion constant of a homotypic pentasaccha-

ride–pentasaccharide interaction in the

presence of calcium will be too low to be

determined by such an approach. Anyway,

these results unequivocally prove the pres-

ence of a speciﬁc interaction between cal-

cium and 1.

Use of paramagnetic ions.—The use of

lanthanide trivalent paramagnetic cations to

mimic the inﬂuence of the calcium ions was

in our hands not so conclusive. Indeed, ei-

ther the chemical shift variations were too

low to be safely interpreted (case of Pr³⁺,

Nd³⁺, Eu³⁺, Yb³⁺, Lu³⁺), or the relaxation

rate variations were quasi-uniform all along

the molecule (case of Gd³⁺). Since, as al-

ready noticed, the cation charge is important

in such phenomena, we have decided to un-

dertake an NMR study with divalent

cations, such as Ni²⁺, Co²⁺and Mn²⁺. As

detailed below, more signiﬁcant effects were

observed with these ions. Although their

ionic radii are slightly smaller than that of

Ca²⁺—respectively 0.69, 0.72 and 0.80 A—

˚

Determination of the interaction sites.—

The important point, which can justify the

use of NMR spectroscopy as a local descrip-

tor of the molecular events, was the determi-

nation of the interaction sites of calcium

with the pentasaccharide. In the literature,

the complexes formed between sugars or

polyols with metal cations are usually triden-

tate and ascribable to hydroxyl groups (al-

though hydroxymethyl [44], intracyclic

[19,45] and interglycosidic oxygen [19] or ni-

trogen [15] can participate to the reaction).

Their formation requires a speciﬁc steric ar-

rangement of the electron donor groups,

namely ax, eq, ax sequence or 3-syn–ax

(rare in sugars) on six-membered rings,

which could help competing efﬁciently with

water molecules for occupancy of the ﬁrst

coordination sphere of a metallic cation. If

the reaction is favourable enough, a change

in conformation can occur to put the com-

plexing groups in these particular conﬁgura-

tions. Nevertheless, the study of structure 1

they represented good candidates for mim-

icking calcium with more important effects

on the NMR spectrum.

Divalent manganese, being a symmetrical

ion, cannot be used as a chemical shift

reagent, as its pseudo-contact contribution

vanishes. In contrast, its high electronic spin

value (5/2) induces large variations in the re-

laxation times of the neighbouring nuclei.

The expressions of the electronic contribu-

tion to nuclear relaxation given by Solomon

[46] and Bloembergen [47] indicate that the

electronic contribution to self-relaxation time

T^e₁is proportional to the distance r_jbetween

the metallic centre and the jth nucleus. It is

thus possible (to a ﬁrst approximation) to

write:

1

T₁^e

=

+c

(5)

T^o₁^bsT^f₁^ree

where c is the concentration of paramagnetic

spin and 1/T^o₁^bsand 1/T^f₁^reethe longitudinal

58

B. Henry et al. / Carbohydrate Research 315 (1999) 48–62

Fig. 4. Superposition of ﬁve structures for 1 determined by molecular modelling under NMR constraints with, for each, the

average location of the two Mn²⁺cations. The NMR data have been recorded in the presence of 11.4 equiv of calcium.

self-relaxation rates measured in the presence

and in the absence of Mn²⁺. Moreover:

shift variations of two protons is consequently

linearly proportional to the relative variation

of Dl_complex. This remark was used to relate

the relative chemical shift variations with the

location of the ion.

e

1/6

ꢂ n

T_1i

r_i

r_j

=

(6)

T^e_1j

Location of Mn²⁺and Co²⁺cations.—An

attempt to ﬁt the experimental data with the

location of a single Mn²⁺ion was unsuccess-

Thus, variation of the carbon longitudinal

self-relaxation rates of the oligosaccharide was

monitored upon addition of MnCl₂.

2

ful. Indeed large values were found, result-

The strong anisotropic magnetic susceptibil-

ity of the Co²⁺or Ni²⁺ions implies that they

constitute good chemical shift reagents. We

have undertaken the study on Co²⁺which

presents a larger ionic radius than Ni²⁺. Dur-

ing cobalt addition, some proton signals were

shifted in the opposite direction relatively to

what was observed in calcium. This represents

an indication that it may be possible to locate

the interaction site(s) through triangulation

methods. However, the spectral lines were

progressively broadened, which shows that

this ion also contributes efﬁciently to trans-

verse relaxation. For this reason, it was im-

possible to reach the plateau value for which

the chemical shifts were no longer affected by

the addition of cobalt, and thus to determine

the equilibrium constant directly. The begin-

ning of the curve representing, for one proton,

the chemical shift variation Dl_obsas a function

of the cobalt concentration [Co], conﬁrmed

that Dl_obsis linearly proportional to the bind-

ing constant K, to [Co], and to the shift in the

pure complex Dl_complex. In the accessible ex-

perimental domain, the ratio of the chemical

ing from the incapability of simultaneously

minimising the variation of the self-relaxation

rates observed on the ﬁrst hand in the Gal^I,

Fuc^II, GlcNAc^IIIand on the second hand in

the Gal^IV, Glc^Vdomain. Therefore, a more

complex procedure involving two Mn²⁺ions

Table 3

Main distances deﬁning the location of Mn^Iand Co^Ia

˚

Distance (A)

Manganese

Cobalt

b

Distance X–X

0.91

2.85

c

Distance Co–Mn

Distance X–Gal^Id

Distance X–Fuc^d

6.692.6

12.290.7

9.890.7

13.390.7

6.692.6

12.091.3

9.691.1

11.091.6

Distance X–GlcNAc^d

^aThe values are computed on the 35 low-energy structures

for 1 in the presence of calcium.

^bAverage distance between the minimised ion location and

the average one.

^cAverage distance between the average locations of the

cobalt ion and of the manganese ion.

^dDistances between the average paramagnetic ion and the

pyranoside unit, computed as the distance between the aver-

age ions X and the pseudo-atom localised at the barycentre

of carbons 1, 3 and 5 of the unit.

B. Henry et al. / Carbohydrate Research 315 (1999) 48–62

59

Fig. 5. One low-energy structure for 1 with 2×100 Mn²⁺ions (blue triangles) and 2×100 Co²⁺ions (orange triangles) for which

coordinates have been determined by the minimisation procedure described in Section 2.

was necessary (see Section 2) as suggested by

the fact that polycations are more likely to be

found in methanol [6].

random drawing of initial ionic coordinates,

the ﬁt, and the clustering procedure (see Sec-

tion 2) fall in this region. Fig. 5 shows one of

the low-energy structures for 1 with some

Mn²⁺ions resulting from our simulation.

For cobalt, as for manganese, it was neces-

sary to introduce a second cobalt ion in order

to ﬁt the data and to explain chemical shift

variations for protons in the lactose part as

well as in the Lewis^Xpart of the molecule. The

resulting locations of both regions are in good

agreement with those found for manganese.

The locations found for the cobalt ion are

however less precise than the manganese sites

even for region I. The average distance be-

tween the minimized ion locations and the

average one X increased by a factor of 3

when passing from Mn²⁺to Co²⁺(Table 3).

Five energy-minimised 3D structures of 1 in

methanol with the regions of highest probabil-

ity for the Mn²⁺ions are displayed in Fig. 4.

Two regions can be discerned: the ﬁrst one,

˚

denoted I, is located at about 10 A from the

nearest atoms of the Lewis^Xmoiety, equidis-

tant from the Fuc and Gal units. Table 3

summarises the principal statistical results ob-

tained for this location of the manganese ion.

The second region, denoted II, is less precise,

but is closer to the lactose part. These results,

obtained on average locations of the man-

ganese ions, were conﬁrmed when considering

a given structure for 1. In this case, all the

accepted locations of Mn^Iobtained after the

60

B. Henry et al. / Carbohydrate Research 315 (1999) 48–62

˚

Region II of cobalt is roughly identical to that

of manganese. For region I, this lack of preci-

sion appears when we compare either the dis-

tribution of the cobalt locations for a given

structure for 1 (Table 4, Fig. 5), or the average

cobalt locations determined for each structure.

This is certainly due to the larger number of

parameters to ﬁt in the case of the cobalt ion.

In our geometry, where the ion is not sur-

rounded by the molecule, a variation of the

ion coordinates can be, at least partially, com-

pensated by a variation of the orientation of

the principal axis of the paramagnetic suscep-

tibility tensor to give the same effect on the

NMR data. Anyway, there is still a site of

interaction near the Lewis^Xmotif, but com-

paratively to the manganese ion, its distance

to the glucosamine unit decreases by about 2

\10.0 A in the case of the manganese. This

tends to prove that the difference of locations

between cobalt and manganese arises from the

inﬂuence of the CO–NH group. This differ-

ence is certainly due to the smaller size of the

Co ion and to different constraints on the

symmetry of the complex.

In Table 4, where the distances between the

average locations of the paramagnetic ions

X and the heavy atoms of the Lewis^Xpart

are reported, it can be observed that essen-

tially the same atoms of the fucose unit are

close to the Co and Mn ions. The effect of the

N-acetyl group also appears, since distances

involving these atoms are found to be short in

the cobalt case. Although it does not consti-

tute a formal proof, the protons whose reso-

nances are the most shifted in the presence of

calcium, are carried by—or are close to—the

heavy atoms of Table 4.

˚

A (Table 3). It is equidistant from the fucose

unit and the methyl group of the glucosamine

unit. In fact, the average distances between

cobalt and the CO–NH group are about 2.5

˚

A shorter than for manganese. The broader

4. Conclusion

distribution of the cobalt coordinates in re-

gion I leads for some PLEX structures to

average distances between cobalt and the

We have shown, using high-resolution

NMR spectroscopy, that a pentasaccharide

containing the Lewis^Xgroup interacts with

different cations. We have been able to deter-

mine the binding constants with calcium in

dimethyl sulfoxide and methanol. The value of

the binding constant in methanol is not

greater than that reported for some fura-

nosides [8]. The expected proportion of 1:100

in the binding constant [22] for complexation

of calcium by methyl furanosides in water and

in methanol allows us to understand why it is

not possible to evidence such an interaction in

water for the pentasaccharide through NMR

spectroscopy. However, a speciﬁc interaction

exists, and the solution structures of this pen-

tasaccharide in methanol in the presence and

in the absence of calcium have been deter-

mined. These molecular structures have al-

lowed the characterisation of the interaction

sites between the pentasaccharide and two

divalent paramagnetic ions¹(Co²⁺and Mn²⁺).

The experimental results indicate that there

˚

CO–NH group of B6.5 A, while it is always

Table 4

Shortest distances between the heavy atoms of the Lewis^X

part and the average location X of the paramagnetic ion^a

Atom

Manganese

Cobalt

b

˚

Distance (A) Rank Distance (A) Rank

Fuc O-3

Fuc O-2

Fuc C-3

Fuc C-2

7.0090.52

7.7190.59

8.2790.54

8.3190.58

8.8890.49

9.3290.49

9.3790.49

9.7890.58

1

2

3

4

5

6

7

8

8.3091.90

6.7191.79

8.8191.81

7.9891.84

4

1

9

2

Fuc O-4

Gal^IC-6

9.3991.82

9.0391.96

8.7691.76

8.2791.77

8.4091.92

8.4291.92

8.6891.73

8.8791.60

9.1691.78

9.7191.62

13

11

8

3

5

6

7

10

12

14

Gal^IO-6

Fuc C-1

GlcNAc C^CO

GlcNAc C^CH

GlcNAc O^CO

GlcNAc N

GlcNAc O-3

GlcNAc C-2

GlcNAc C-3

3

¹The Cartesian coordinates of a low-energy structure for 1

and the associated best-ﬁt locations for the Mn²⁺and Co²⁺

ions are available upon request from the authors.

a

˚

Only the distances shorter than 10 A are reported.

^bRanks: distances in increasing order.

B. Henry et al. / Carbohydrate Research 315 (1999) 48–62

61

[10] S.J. Angyal, J.A. Mills, Aust. J. Chem., 32 (1979) 1993–

2001.

[11] S.J. Angyal, D.C. Craig, Carbohydr. Res., 241 (1993)

1–8.

[12] F. Cadet, Spectrosc. Lett., 29 (1996) 1353–1365.

[13] P. Ronge`re, N. Morel-Desrosiers, J.-P. Morel, J. Chem.

Soc., Faraday Trans., 91 (1995) 2771–2777.

[14] A. Vesala, R. Ka¨ppi, Polyhedron, 4 (1984) 1047–1050.

[15] M. Miyazaki, S. Nishimura, A. Yoshida, N. Okubo,

Chem. Pharm. Bull., 27 (1979) 532–535.

[16] K. Dill, D. Carter, Ad6. Carbohydr. Chem. Biochem., 47

(1989) 125–166.

[17] G. Siuzdak, Y. Ichikawa, T.J. Caulﬁeld, B. Munoz,

C.-H. Wong, K.C. Nicolaou, J. Am. Chem. Soc., 115

(1993) 2877–2881.

[18] S. Penn, M. Cancilla, C. Lebrilla, Anal. Chem., 68 (1996)

2331–2339.

[19] G. Siuzdak, Z.-L. Zheng, J.Y. Rhampal, Y. Ichikawa,

K.C. Nicolaou, F.C.A. Gaeta, K.S. Chatman, C.-H.

Wong, Bioorg. Med. Chem. Lett., 4 (1994) 2863–2866.

[20] P. Brocca, P. Berthault, S. Sonnino, Biophys. J., 74

(1998) 309–318.

[21] M.R. Wormald, C.J. Edge, R.A. Dwek, Biophys.

Biochem. Res. Commun., 180 (1991) 1214–1221.

[22] H. Lo¨nnberg, A. Vesala, R. Ka¨ppi, Carbohydr. Res., 86

(1980) 137–142.

[23] H. Desvaux, P. Berthault, N. Birlirakis, M. Goldman,

C.R. Acad. Sci. Paris, Se´r. II, 317 (1993) 19–25.

[24] H. Desvaux, P. Berthault, N. Birlirakis, Chem. Phys.

Lett., 233 (1995) 545–549.

[25] P. Berthault, N. Birlirakis, G. Rubinstenn, P. Sinay¨, H.

Desvaux, J. Biomol. NMR, 8 (1996) 23–35.

[26] I. Lundtiand, C. Pedersen, Acta Chem. Scand., Ser. B, 30

(1976) 680–684.

is a relatively well-deﬁned region of interac-

tion close to the Lewis^Xmotif, and a less

speciﬁed area in the lactose part of the

molecule. The results obtained for these two

ions are similar but not identical; this is cer-

tainly due to the difference of ionic radius and

of symmetry of the complexes. The extension

to the structure of the calcium complex is

obviously not straightforward since these two

parameters are different, and the chemical

shift variations observed during the formation

of the complex result from a diamagnetic ef-

fect. However, the similarity of the cobalt and

manganese results and the remark that the

protons for which chemical shifts are more

affected by the addition of calcium (Table 2)

are those whose carbons are found to be close

to the paramagnetic ions (Table 4), make it

reasonable to think that the site of interaction

between Ca²⁺and molecule 1 is located in the

vicinity of the fucose unit. These experimental

results substantiate Hakomori’s hypothesis on

the structure of the complex between the

Lewis^Xand divalent ions, involving homotypic

interactions between oligosaccharides.

[27] R.K. Jain, K.L. Matta, Carbohydr. Res., 226 (1992)

91–100.

[28] H. Lo¨nn, Carbohydr. Res., 139 (1985) 105–113.

[29] G.H. Veeneman, S.H. Van Leeuwen, J.H. Van Boom,

Tetrahedron Lett., 31 (1990) 1331–1334.

[30] T. Ehara, A. Kameyama, Y. Yamada, H. Ishida, M.

Kiso, A. Hasegawa, Carbohydr. Res., 281 (1996) 237–

252.

[31] J. Dahmen, G. Gnosspelius, A.-C. Larsson, T. Lave, G.

Noori, K. Palsson, Carbohydr. Res., 138 (1985) 17–28.

[32] L. Mu¨ller, J. Am. Chem. Soc., 101 (1979) 4481–4484.

[33] A. Bax, R.H. Griffey, B.L. Hawkins, J. Magn. Reson., 55

(1983) 301–315.

Acknowledgements

This work was partially funded by a grant

from the Centre National de la Recherche

Scientiﬁque (CNRS), ‘Actions Coordonne´es

Concerte´es des Sciences du Vivant (ACC-SV5:

Interface Chimie-Physique-Biologie: Biologie

structurale et pharmacologie)’.

[34] A. Bax, M.F. Summers, J. Am. Chem. Soc., 108 (1986)

2093–2094.

[35] A.T. Bru¨nger, ^X-PLOR Version 3.1: a System for X-ray

Crystallography and NMR, Yale University Press, New

Haven, 1992.

[36] W.H. Press, W.T. Vetterling, B.P. Flannery, Numerical

Recipes in C. The Art of Scientiﬁc Programming, Cam-

bridge University Press, Cambridge, 1988.

[37] A. Vesala, H. Lo¨nnberg, R. Ka¨ppi, J. Arpalahti, Carbo-

hydr. Res., 102 (1982) 312–315.

[38] J. Reuben, J. Phys., 79 (1975) 2154–2157.

[39] L. Banci, I. Bertini, C. Luchinat, Nuclear and Electron

Relaxation, VCH, Weinheim, 1991.

References

[1] S.I. Hakomori, Pure Appl. Chem., 63 (1991) 473–482.

[2] D.S. Spillmann, M.M. Burger, J. Cell. Biochem., 61

(1996) 562–568.

[3] N.V. Bovin, Biochemistry (Moscow), 61 (1996) 694–704.

[4] I. Eggens, B. Fenderson, T. Toyokuni, B. Dean, M.

Stroud, S.I. Hakomori, J. Biol. Chem., 264 (1989) 9476–

9484.

[5] N. Kojima, B.A. Fenderson, M.R. Stroud, R.I. Gold-

berg, R. Habermann, T. Toyokuni, S.I. Hakomori, Gly-

coconjugate J., 11 (1994) 238–248.

[6] J.A. Rendleman Jr., J. Org. Chem., 31 (1966) 1839–1845.

[7] S.J. Angyal, Tetrahedron, 30 (1974) 1695–1702.

[8] S.J. Angyal, Ad6. Carbohydr. Chem. Biochem., 47 (1989)

1–43.

[40] G. Rubinstenn, P. Sinay

9 , P. Berthault, J. Phys. Chem. A,

101 (1997) 2536–2540.

[41] H. Desvaux, P. Berthault, N. Birlirakis, M. Goldman, J.

Chim. Phys., 91 (1994) 646652.

[42] M.C.R. Symons, J.A. Benbow, H. Pelmore, J. Chem.

Soc., Faraday Trans. 1, 78 (1982) 3671–3677.

[9] J. Mills, Biochem. Biophys. Res. Commun., 6 (1961) 418–

421.

62

B. Henry et al. / Carbohydrate Research 315 (1999) 48–62

[43] M.C.R. Symons, J.A. Benbow, H. Pelmore, J. Chem.

Soc., Faraday Trans. 1, 80 (1984) 2017–2026.

[44] S.J. Angyal, D. Greeves, V.A. Pickles, Carbohydr. Res.,

35 (1974) 165–173.

[45] S.J. Angyal, D. Greeves, L. Littlemore, V.A. Pickles,

Aust. J. Chem., 29 (1976) 1231–1237.

[48] M. Piotto, V. Saudek, V. Sklenar, J. Biomol. NMR, 2

(1992) 661–665.

[49] V. Sklenar, M. Piotto, R. Leppik, V. Saudek, J. Magn.

Reson., 102A (1993) 241–245.

[50] S. Mori, J.M. Berg, P.C.M. van Zijl, J. Biomol. NMR, 7

(1996) 77–82.

[46] I. Solomon, Phys. Re6., 99 (1955) 559–565.

[47] N. Bloembergen, J. Chem. Phys., 27 (1957) 572–573.

[51] H. Desvaux, P. Berthault, N. Birlirakis, M. Goldman,

M. Piotto, J. Magn. Reson. A, 113 (1995) 47–52.

.

Article Doi

NMR study of a Lewis(X) pentasaccharide derivative: Solution structure and interaction with cations

DOI: 10.1016/S0008-6215(98)00301-2

Source and publish data:

Authors:

Article abstract of DOI:10.1016/S0008-6215(98)00301-2

Full text of DOI:10.1016/S0008-6215(98)00301-2

Products guided by the article

R&D Labs maybe for 227749-87-3

Relevant to this article

Hot Product