Towards targeted MRI: New MRI contrast agents for sialic acid detection

Article abstract of DOI:10.1002/chem.200400369

The detection of sialic acid in living systems is of importance for the diagnosis of several types of malignancy. We have designed and synthesized two new lanthanide ion ligands (L¹ and L²) that are capable of molecular recognition o

Full text of DOI:10.1002/chem.200400369

FULL PAPER
Towards Targeted MRI: New MRI Contrast Agents for Sialic Acid Detection
Luca Frullano,[a, b] Jan Rohovec,[c] Silvio Aime,[d] Thomas Maschmeyer,[a, e]
[f]
M. Isabel Prata,[f] J. J. Pedroso de Lima, Carlos F. G. C. Geraldes,[g] and
Joop A. Peters*[a]
Abstract: The detection of sialic acid in
living systems is of importance for the
diagnosis of several types of malignan-
cy. We have designed and synthesized
two new lanthanide ion ligands (L1 and
L2) that are capable of molecular rec-
ognition of sialic acid residues. The
basic structure of these ligands consists
of a DTPA-bisamide (DTPA, diethyl-
enetriamine pentaacetic acid) whose
amide moieties each bear both a bor-
onic function for interaction with the
diol groups in the side chain of sialic
acid, and a functional group that is pos-
itively charged at physiologic pH
values and is designed to interact with
the carboxylate anion of sialic acid.
The relaxometric properties of the
Gd3+ complexes of these two ligands
were evaluated. The relaxivity of the
GdL1 complex has a significant second-
sphere contribution at pH values above
the pKa of its phenylboronic acid
moiety. The interaction of the Gd3+
complexes of L1 and L2 with each of
several saccharides was investigated by
assay. The results show that both com-
plexes recognize sialic acid with good
selectivity in the presence of other
sugars. The adduct formed by GdL2
with sialic acid has the higher condi-
tional formation constant (50.43ꢀ
4.61mꢁ1 at pH 7.4). The ability of such
complexes to recognize sialic acid was
confirmed by the results of a study on
the interaction of corresponding radio-
means of
a
competitive fluorescent
labeled complexes
(
153SmL1 and
153SmL2) with C6 glioma rat cells.
153SmL2 in particular is retained on the
cell surface in significant amounts.
Keywords: borates · carbohydrates ·
drug design · lanthanides · magnetic
resonance imaging
Introduction
the contrast depends essentially on three factors, 1) the
water proton density, 2) the longitudinal relaxation time, T1,
and 3) the transverse relaxation time, T2, of these pro-
tons.[1–4] Paramagnetic contrast agents (CAs) are often ad-
ministered prior to the scan to improve the contrast in the
image. The role of these CAs, usually Gd3+ complexes, is to
shorten the relaxation times of the water 1H NMR reso-
The excellent contrast and spatial resolution of the images
obtained by magnetic resonance imaging (MRI), particularly
those of soft tissues, have made this technique one of the
fastest growing diagnostic methodologies. An MRI image is
generated from the NMR resonance of water protons and
[a] Dr. L. Frullano, Prof. Dr. T. Maschmeyer, Dr. J. A. Peters
Laboratory of Applied Organic Chemistry and Catalysis
Julianalaan 136, 2628 BL Delft (The Netherlands)
Fax : (+31)15-278-4289
[e] Prof. Dr. T. Maschmeyer
School of Chemistry
The University of Sydney
NSW 2006 (Australia)
E-mail: j.a.peters@tnw.tudelft.nl
[f] Dr. M. I. Prata, Prof. Dr. J. J. P. de Lima
Department of Biophysics of the Faculty of Medicine
University of Coimbra
[b] Dr. L. Frullano
Chemistry Department
Northwestern University
Evanston, Illinois 60208-3113 (USA)
3000 Coimbra (Portugal)
[g] Prof. Dr. C. F. G. C. Geraldes
Departamento de CiÞncias e Tecnologica e
Centro NeurociÞncias
[c] Dr. J. Rohovec
Department of Chemistry
Universita Karlova
Universidade de Coimbra
12840 Prague (Czech Republic)
3049 Coimbra (Portugal)
[d] Prof. S. Aime
Dipartimento di Chimica I.F.M.
Università di Torino
10125 Torino (Italy)
Chem. Eur. J. 2004, 10, 5205 – 5217
DOI: 10.1002/chem.200400369
ꢀ 2004 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
5205

FULL PAPER
nance. CAs currently on the
market suffer from several dis-
advantages, the most important
of which are poor efficiency
with regard to the shortening of
T1 and T2, and a lack of specific-
ity. Once administered, these
agents rapidly equilibrate non-
specifically between the intra-
vascular and the interstitial
compartments. For this reason,
a great deal of attention has
been dedicated to the develop-
ment of new and more efficient
contrast agents.[5,6] The delivery
of higher doses of CAto the
target site through the exploita-
tion of receptors or molecular
determinants would improve
visualization of abnormalities
and the CAcould be used to
obtain not only anatomical in-
formation but also information
at the molecular level. Molecu-
lar imaging would make it possi-
ble to study and understand cel-
lular events or to monitor gene
delivery and expression in gene
therapy.[7–10]
Scheme 1. Molecular structures of a Neu5Ac end group in a glycoprotein or glycolipid (R) and the targeting li-
gands L1 and L2.
Amajor difficulty in the development of targeted CAs is
the very low concentration of receptors (10ꢁ9–10ꢁ13 molgꢁ1)
in tissues, combined with the low intrinsic sensitivity of
MRI.[11] Several approaches have been applied to this prob-
lem and some progress has been made by using monoclonal
antibodies (mAbs) covalently conjugated with MRI CAs.
However, modification of mAbs can alter their affinity to-
wards the target as well as their pharmacokinetic proper-
ties.[4] There is thus considerable interest in the development
of alternative targeting strategies.
Receptors often consist of carbohydrates on glycoproteins
or glycolipids, and in many cases sialic acids are involved.
The sialic acids are a family of C9 monosaccharides with a
carboxy group at the anomeric carbon atom (pKa =2.2) that
gives the molecule a negative charge at physiological pH
(Scheme 1). These compounds frequently occur as the termi-
nal residue of a glycan chain in a glycoprotein or glycolipid
present on the cell surface or in an intracellular membrane
(e.g. Golgi apparatus). The biological roles played by sialic
acids are the result of the physicochemical properties of
these molecules (size, hydrophilicity, and negative charge at
physiological pH) and of their exposed positions in glycan
chains. They appear to be involved in the binding and trans-
port of positively charged ions or molecules and in the at-
traction and repulsion of cells and molecules. Sialic acids
can also act as a protective shield for the subterminal por-
tion of a molecule, but their most important functions con-
cern cellular and molecular recognition.[12–16]
lar, a correlation between an altered sialylation pattern and
lowered defensive activity against viral infections and tumor
cells has been reported.[17] Lemieux et al. took advantage of
the overexpression of sialic acid on tumor cell surfaces (up
to 109 sialic acid residues per tumor cell as compared to
only 20”106 for a normal human erythrocyte) to detect
tumor cells.[18] By supplying N-levulinoylmannosamine, an
unnatural substrate for the sialoside biosynthetic pathway,
to Jurkat cells these authors were able to convert the sub-
strate into N-levulinoyl sialic acid, a ketone-substituted
sialic acid. This keto group can be used as an anchor for co-
valent binding of an aminoxy-substituted GdDTPA(DTPA,
diethylenetriamine pentaacetic acid), which results in con-
trast enhancement and may make the molecule observable
in an MRI image.
Herein, we report the results of studies on two Gd3+ com-
plexes (GdL1 and GdL2) built with ligands containing DTPA
and phenylboronate functions (see Scheme 1). These com-
plexes were designed to interact reversibly with sialic acid.
The advantage of this approach is that the contrast agent is
in a dynamic equilibrium between its sialic-acid-bound and
free states. The free complex can eventually be excreted by
the normal physiological mechanism, which shifts the equi-
librium away from the bound form.
Many efforts have been made to develop biomimetic
sugar receptors that operate in water.[19,20] However, the
only promising class of sugar receptors developed so far is
that of the arylboronic acids, which are not biomimetic.[21]
The ability of boric and phenylboronic acid to interact with
sugars has been known for a long time.[22–24] Phenylboronic
Sialic acid concentration has been proposed as a prognos-
tic and diagnostic indicator for several diseases. In particu-
5206
ꢀ 2004 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chemeurj.org
Chem. Eur. J. 2004, 10, 5205 – 5217

MRI Contrast Agents
5205 – 5217
acid is a weak acid (pKa =8.67)[25] that bonds covalently and
reversibly with 1,2- or 1,3-diols to give five- or six-mem-
bered cyclic esters (see Scheme 2). The stability of the esters
formed by the boronate anion (tetragonal) is orders of mag-
nitude higher than that of the esters formed by boronic acid
(trigonal).[26] The binding ability of boronic acid is therefore
optimal under basic conditions (pH>8–9) since the concen-
tration of the boronate anion is maximal in such an environ-
ment.
acid/Neu5Ac adduct formation is higher than those for vari-
ous other monosaccharides commonly present in glycan
chains on cell surfaces. Free Neu5Ac differs from sialic acid
residues in glycan chains; in a glycan chain, the C2 position
of the acid is involved in the glycosidic linkage, which pre-
cludes interaction with phenylboronic acid at the C1/C2 po-
sition. Consequently, it is reasonable to expect a weaker in-
teraction between phenylboronic acid and sialic acids in gly-
coconjugates than between the same compound and free
Neu5Ac.
To increase the specificity and stability of the interaction
between the target receptor and sialic acid, we introduced
another functional group to our artificial receptors in addi-
tion to phenylboronic acid. The role of this additional group
is to interact with the carboxy group at the C1 position of
sialic acid (Scheme 1). Guanidinium cations are very weakly
acidic (pKa ꢂ12.5) and have the capacity to form intermo-
lecular contacts mediated by H-bonding interactions. The
guanidinium ion is present in many natural products and is
often involved in molecular recognition.[31] For example, in
E-selectin a guanidinium ion on an arginine residue is the
binding site for the carboxylate group of sialyl Lewis X
acids.[32] This example inspired us to include an aminoimida-
zolium group in the design of L1 to allow binding of the car-
boxylate group of Neu5Ac.
Compound L2 was designed to interact with the carboxy
group at the C1 position of Neu5Ac through an aminometh-
yl group present in the meta-position of L2 relative to the
boronic function. This aminomethyl moeity is protonated
under physiological conditions. The design of both ligands
was completed by inclusion of a central chelating unit based
on the DTPAstructure, which is known to afford a stable
coordination cage for lanthanide ions (see Scheme 1).[1–4]
Molecular modeling studies indicate that both L1 and L2 can
occur in conformations that allow cooperative two-site bind-
ing of sialic acid through 1) ester formation by interaction of
the boronate function on the ligand with the geminal diol
function at C8/C9 of Ne5Ac, and 2) an electrostatic interac-
tion between the positively charged moiety on the ligand
and the carboxylate group of Neu5Ac.
Scheme 2. The equilibria involved in a diol–boronic acid interaction.
The requirement of basic conditions for optimal interac-
tion is a drawback for the use of these moieties in saccha-
ride receptors. Wulff has shown that the conversion of a
trigonal boronic into a tetragonal boronate function occurs
at a much lower pH value when an aminomethyl group is in-
troduced at the ortho-position (Scheme 3).[27,28] This effect
The two ligands L1 and L2 were synthesized and the re-
laxometric properties of their Gd3+ complexes were evaluat-
ed. Interaction of the ligands with Neu5Ac and other com-
peting monosaccharides present in glycan chains was investi-
gated by means of a three-component competitive fluores-
cence assay developed by Wang and Springsteen.[26] We ex-
ploited the excellent sensitivity of radiometric methods to
investigate the interaction of the 153Sm complexes of the li-
gands with C6 rat glioma cells. It has been shown that these
cells have high membrane concentrations of sialic acid.[33]
Scheme 3. The pH-dependent equilibria characteristic of an o-aminome-
thylphenylboronic acid.
can be ascribed to a B–N interaction in the phenylboronate
species that results in a decrease in the pKa value of the ter-
tiary amino group to around 5, and an increase in that of
the boronic function to about 12.[29] The presence of the tet-
ragonal species over a broader pH range favors the interac-
tion of this type of compound with saccharides at physiolog-
ical pH levels.
We previously studied the interaction of phenylboronic
acid with Neu5Ac, the most widespread form of sialic acid
(Scheme 1).[30] The boronic group appeared to interact both
with the glycerine moiety at C6 to form five- and six-mem-
bered cyclic esters, and with the carboxy group and the hy-
droxy group at C2 to form a five-membered cyclic ester.
The formation constant we determined for phenylboronic
Results and Discussion
Synthesis: The ligands prepared for this study, L1 and L2, are
both composed of
a central lanthanide-chelating unit
flanked by two identical groups designed to interact with
sialic acid. Ligand L1 was synthesized from tris-(2-amino-
ethyl)amine (TREN). Two of the three amine functions of
Chem. Eur. J. 2004, 10, 5205 – 5217
www.chemeurj.org
ꢀ 2004 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
5207

J. A. Peters et al.
FULL PAPER
this compound were protected with tert-butoxycarbonyl
(Boc) groups by the procedure of Hamdaoui et al.,[34] which
gave compound 1 (Scheme 4). Treatment of 1 with 2-methyl-
thio-2-imidazoline hydroiodide in refluxing ethanol, fol-
lowed by deprotection with HCl afforded 3 in excellent
yield. The boronic acid function was introduced by reductive
amination of 3 with 2-formylphenylboronic acid and sodium
borohydride, which led to a mixture of mono- and bis-substi-
tuted derivatives of 3. Separation of these derivatives was
achieved by ion exchange chromatography on a weakly
acidic cation exchanger (Amberlite CG50), with gradient
elution in aqueous ammonia. Ligand L1 was obtained by
condensation of 4 with DTPA-bisanhydride in ethanol.
The bis-Boc-protected precursor of ligand L2 (5) was pre-
pared by a published procedure.[35] Deprotection by treat-
ment with TFAgave 6 almost quantitatively. Ligand L2 was
obtained by reductive amination of 6 with excess 3-formyl-
phenylboronic acid and sodium borohydride.
an inflection point at about pH 4.9. This shift change marks
the transition between a trigonal and a tetragonal boron
ꢁ
atom that results from formation of a B N bond upon de-
protonation of the amino group (see Scheme 3). By fitting
the chemical shift titration curve to the appropriate equa-
tions, we determined the pKa value of the amino group to
be 4.93. This value is considerably lower than the pKa of
benzylamine (pKa =9.40ꢀ0.08, m=0.1),[36] which demon-
strates that the ammonium group is considerably stabilized
by the presence of the ortho-boronic function. It was not
possible to determine the pKa value of the boronic function
from the titration curve since no other jump in chemical
shift occurred. This result indicates that the pKa value of the
boronic function is higher than 12. The observed pKa trends
agree well with those reported by Wiskur et al. for o-amino-
methylphenylboronic acid.[29]
The situation is different for L2 because the aminomethyl
group is in the meta-position with respect to the boronic
function and B–N interaction is therefore impossible. The
11B NMR chemical shift change (9.4 to ꢁ16.6 ppm) with in-
creasing pH value can, once again, be ascribed to the con-
version of the boron functionality from a trigonal into a tet-
11B NMR shift titrations: The chemical shift of the 11B reso-
nance of ligand L1 was recorded as a function of the pH
value and was found to jump from 9.4 to ꢁ11.4 ppm, with
Scheme 4. Preparation of ligands L1 and L2. i) 2-methylthio-2-imidazoline hydroiodide, EtOH, 5 h, reflux; ii) HCl 37%, EtOH, 1.5 h, RT, 87%; iii) 2-for-
mylphenylboronic acid, MeOH/TEA(3:1 v/v), 6 h, RT; iv) NaBH 4, 12 h, RT, 26%; v) DTPA-bisanhydride, zeolite KA, EtOH, 6 h, 95%; vi) TFA, 12 h,
RT, 97%; vii) 3-formylphenylboronic acid, MeOH/TEA(2:1 v/v), 2 h, RT; viii) NaBH , MeOH/TEA (2:1 v/v), 12 h, RT, 90%. TEA, triethylamine; TFA,
4
trifluoroacetic acid.
5208
ꢀ 2004 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chemeurj.org
Chem. Eur. J. 2004, 10, 5205 – 5217

MRI Contrast Agents
5205 – 5217
ragonal geometry (Scheme 3). In this case, this transition is
the result of the reaction of the boron function with an OHꢁ
ion to form a boronate. The pKa value concerned was deter-
mined to be 7.96, which is somewhat lower than that report-
ed for phenylboronic acid (pKa =8.7),[25] probably as a result
of the electron-withdrawing effect of the aminomethyl
group. We conclude that L1 is present in the tetragonal form
over a wider pH range than L2.
Hamiltonian, and ws and wI are the electron and proton
Larmor frequencies, respectively.
At pH 4.5 and 298 K, the relaxivities (measured at
20 MHz) of GdL1 and GdL2 are 5.7 and 4.7 sꢁ1 mmꢁ1, respec-
tively. These values are similar to those reported for Gd3+
complexes of other DTPA-bisamides.[38,39] The pH depen-
dence of the relaxivities of the complexes increases signfi-
cantly at low and high pH values (Figure 1). At pH<2 both
The 11B NMR chemical shift of L1 under basic conditions
is ꢁ11.4 ppm, and that of L2 is ꢁ16.6 ppm. This difference
reflects the different electronic environments of the boron
atoms in the two ligands and confirms the presence of a
B–N interaction in L1, and the absence of such an interac-
tion in L2.
Relaxometric characterization of the Gd3+ complexes: The
Gd3+ complexes of L1 and L2 were synthesized from equi-
molar amounts of GdCl3 and ligand in water at room tem-
perature. The pH value was maintained at 5.5–6.5 during
this synthesis.
The efficiency of an MRI CAis generally expressed in
terms of relaxivity (rip in sꢁ1 mmꢁ1), a parameter that indi-
cates the ability of the CAto shorten the relaxation times of
solvent water protons. Relaxivity is usually considered to be
the result of a combination of inner-sphere (ri1s) and outer-
sphere contributions (ro1s), which arise from the water mole-
cules directly coordinated to the metal center and those
freely diffusing around it, respectively.[1–4] Sometimes, an ad-
ditional contribution originating from second-sphere water
molecules can be identified. These water molecules are not
directly coordinated to the metal ion but are bound to the
complex through hydrogen bonds or other relatively weak
interactions.
The inner-sphere contribution is described by Solomon–
Bloembergen–Morgan theory [Equations (1)–(4)].[37]
½CŠ Á q
ri1s
¼
ð1Þ
ð2Þ
1000 Á 55:56 Á ðT1M þ tMÞ
Figure 1. pH dependence of the relaxivity of a) GdL2 and b) GdL1 at
298 K and 20 MHz. The curves shown for pH>2 are the result of fitting
the data to Equations (5)–(7).
D
ðT1MÞꢁ1
/
is JðwI,ws,tciÞ
r6
ꢁ1
ðtciÞꢁ1 ¼ ðtMꢁ1 þ tꢁr 1 þ t Þ
ð3Þ
ð4Þ
is
ðtisÞꢁ1 / KiD2Jðws, tvÞ
complexes show a steep increase in relaxivity to about
13 sꢁ1 mmꢁ1, which is comparable to the relaxivity of the
Gd3+ aquo ion. This behavior can be ascribed to decomplex-
ation of the Gd3+ ion. At the other end of the profile, the
relaxivity of GdL2 begins to rise at pH 5–6 and the curve
shows a jump in relaxivity centered around the pKa of the
boron functionality. In contrast, the relaxivity of GdL1 only
begins to rise when a pH value of 8–9 is reached. In both
cases, the increase in relaxivity observed in the basic pH
region (pH>9) is most likely to be caused by activation of
the prototropic exchange mechanism.[39]
[C] is the concentration of the Gd3+ complex, q is the
number of water molecules in the first coordination sphere
of the metal ion, T1M is their proton relaxation time, r is the
Gd3+–Hwater distance, J is the spectral density function, tci
(i=1, 2) are the overall correlation times for the dipolar nu-
cleus–electron interaction, tM is the mean residence lifetime
of the water molecules in the inner coordination sphere, tr is
the reorientational correlation time of the GdHwater vector,
tis is the averaged relaxation time of the Gd3+ unpaired
electrons, Dis and Ki are constants related to the nuclear and
electron relaxation mechanisms, respectively, D2 is the
square of the transient zero field splitting (ZFS) energy, tv is
the correlation time for the processes modulating the ZFS
DTPA-bisamides generally have relatively long exchange
lifetimes, tM. The relaxivity of a complex of this size with a
fast water exchange rate is usually governed mainly by
changes in the rotational correlation time. However, when
Chem. Eur. J. 2004, 10, 5205 – 5217
www.chemeurj.org
ꢀ 2004 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
5209

J. A. Peters et al.
FULL PAPER
the water exchange becomes sufficiently slow, tM starts to
play a role as well. Any phenomenon that speeds up the ex-
change of the water protons between the complex and the
bulk solvent results in an increase in the relaxivity.[1–4]
catalyzed both by H+ and by OHꢁ ions. It has been reported
that the prototropic exchange rate is modulated by the over-
all charge of the complex under basic conditions; exchange
is faster for positively charged and slower for negatively
charged complexes.[39] This effect has been explained in
terms of the ease of access of OHꢁ ions to the coordinated
water molecule. It is likely that the activation of prototropic
exchange plays a role in the relaxivity enhancement ob-
served for the two complexes GdL1 and GdL2 under basic
conditions.
To evaluate the (whole) water exchange rates of GdL1
and GdL2, we determined the transverse 17O relaxation rate
for water in samples containing the complexes as a function
of the temperature (Figure 2). An optimal fit of the experi-
Comparison of the pH dependence of the relaxivities of
GdL1 and GdL2 with the profiles of other GdDTPA-bisa-
mide complexes reveals that the relaxivity of GdL2 begins to
increase at an unusually low pH value.[39,43] The presence of
a jump in relaxivity around the pKa value of the boronic
functions of GdL2 suggests that this phenomenon is related
to the formation of the negatively charged boronate group.
Inspection of molecular models reveals that this complex
may easily adopt a conformation in which the boronate
function is in close proximity to the Gd3+ ion. Therefore, a
possible explanation for this jump in relaxivity is the pres-
ence of second-sphere water molecules near the negatively
charged boronate group. These water molecules may be hy-
drogen bonded to the boronate function. The boronic
groups in GdL1 have a substantially higher pKa value (>12)
than those in GdL2. The boronic acid moiety in GdL1 has no
overall charge in the pH range investigated and is much less
effective at attracting second-sphere water molecules.
The pH profiles of the two complexes were fitted with
Equation (6), which is derived from Equations (2) and (5).
In these equations, kex is the exchange rate of the bulk water
and kexb is the rate of prototropic exchange catalyzed by hy-
droxy ions. Terms accounting for the contribution made by
second-sphere and outer-sphere water molecules (ro1s) are in-
cluded. The concentration of free boronate functions, Cb is
given by Equation (7), which can be derived from the equi-
librium constant of the boric acid–boronate equilibrium and
the appropriate mass balances.
Figure 2. Temperature dependence of the 17O transverse relaxation rate
of a solution of GdL1 (18.7 mm) at pH 6.0 (&), and of a solution of GdL2
*
*
(25 mm) at pH 4.3 ( ) and at pH 12.0 ( ).
mental data with the Swift–Connick equations[40,41] was ob-
tained for the tM values compiled in Table 1. The exchange
lifetime of water in a solution of GdL2 was found to be 25%
Table 1. Water exchange lifetimes (tM) of GdL1 and GdL2, determined
from water 17O NMR relaxation rate (1/T2) measurements. Standard de-
viations are given in parentheses.
Complex
pH
tM [ms]
GdL1
GdL2
GdL2
6.0
4.3
12.0
3.0 (0.2)
4.1 (0.3)
2.9 (0.2)
tM ¼ ðkex þ kexb½OHꢁŠÞꢁ1
ð5Þ
Ctot
q
greater at pH 4 than under basic conditions. This difference
may be ascribed to the change in the overall charge of the
complex from +2 at acidic pH values to ꢁ2 under basic
conditions. Agreater overall negative charge favors the dis-
sociative process by which water molecules leave the com-
plex and thus accelerates water exchange.[42] However, the
observed change in the water exchange rate is not sufficient
to account for the increased relaxivity observed at basic pH
values (Figure 1). This was confirmed by a simulation of the
longitudinal relaxation of a water molecule at 20 MHz pro-
duced from the tM value obtained from the 17O NMR mea-
surements (see Table 1), with all other parameters set to
those of GdDTPA-bismethylamide (GdDTPA-BMA).
obs
1p
r
¼
1000 Á 55:56 Á ½T1M þ 1=ðkex þ kexb½OHꢁŠÞŠ
ð6Þ
Cbw
þ
þ ro1s
00
1000 Á 55:56 Á T
1M
Kb Á Ctot
Cb
¼
ð7Þ
2
Kb þ ½HþŠ
Ctot is the molar concentration of the paramagnetic com-
plex, Cb is the molar concentration of the complex in the
boronate form, T1M is the proton relaxation time of the coor-
dinated water molecule, q is the number of inner-sphere
water molecules, w is the number of second-sphere water
molecules in the boronate form of the complex, and T is
00
The pH profiles of the relaxivities of GdDTPA-bisamides
generally show a sudden increase in relaxivity at basic pH as
a result of the activation of prototropic exchange, which re-
1M
their proton relaxation time. We assumed that the residence
time of these water molecules is very short compared to T100
00
moves the quenching effect of the slow (whole) water ex-
(t0M0 !T ). We disregarded any contribution from exchange
1M
[39]
change on the relaxivity.
Prototropic exchange can be
of the amide protons and the contribution of the amino
5210
ꢀ 2004 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chemeurj.org
Chem. Eur. J. 2004, 10, 5205 – 5217

MRI Contrast Agents
5205 – 5217
groups in the two ligands to the formation of a second-
sphere hydration shell since these functional groups are rela-
tively far from the paramagnetic center. An excellent fit was
obtained for the pH profile of the relaxivity of GdL1 when
relaxation dispersion; NMRD) and fitting this experimental
data to the values calculated by using the Solomon–Bloem-
bergen–Morgan theory.[44–46] The large number of parame-
ters involved in this model makes it advisable to determine
as many of them as possible by independent techniques.
These values can then be fixed during fitting of the NMRD
profile. The NMRD profiles of GdL1 and GdL2 were record-
ed at 298 K and pH 6 and are shown in Figure 3. The
the number of second-sphere water molecules (w) was fixed
00
at zero. For GdL2, an optimal fit was obtained with w/T
=
1M
8.48”104 sꢁ1. The curves obtained are shown in Figure 1 and
the best-fit values for the prototropic exchange rate con-
stants of the two ligands are listed in Table 2, together with
Table 2. Prototropic exchange rates determined from the best-fit pH-de-
pendence curves of the relaxivities of GdL1 and GdL2 at 20 MHz and
258C.[a] The exchange rates of Gd3+ complexes of other DTPA-bisamides
are included for comparison.
Ligand
kexb [sꢁ1 mꢁ1
]
Reference
L1
1.41”1010
5.68”109
4.6”108
this work
this work
[39]
[39]
[39]
L2
7[b]
8[b]
9[b]
8.6”1010
1.2”1010
[a] See the main text for details. [b] The structures of ligands 7–9 are
shown in Scheme 5.
literature values for other relevant Gd3+ complexes of
DTPA-bisamides (Scheme 5).[39] The prototropic exchange
rate constants obtained for GdL1 and GdL2 are in good
agreement with the general trend of increasing prototropic
exchange rate with increasing overall complex (positive)
charge.
2
Figure 3. NMRD profiles of GdL1
(
(
~
), GdL
*
), and GdDTPA-BMA
*
(
), measured at 298 K.
NMRD profile of GdDTPA-BMA is included in this figure
for comparison. At the pH value tested, the relaxivities of
GdL1 and GdL2 are not significantly influenced by catalysis
of the prototropic exchange or by the presence of second-
sphere water molecules. We therefore assumed that the resi-
dence time of the water protons is determined by the
“whole water” exchange rate, which was measured inde-
pendently from the water 17O transverse relaxation rates
(see above). The experimental data were fitted to the theo-
retical values obtained from the Solomon–Bloembergen–
Morgan theory. At low field, the relaxivities of Gd3+ com-
plexes are mainly determined by the electronic relaxation
time, ts. The relatively low relaxivities of GdL1 and GdL2 at
low field compared to those of DOTA-like Gd3+ complexes
under the same conditions (r1 ꢂ12 sꢁ1 mmꢁ1) are indicative of
short electronic relaxation times, as observed for other acy-
clic ligands with low symmetry.[43] By using standard param-
eter values for a (distance of closest approach of a second-
sphere water molecule, 3.8”10ꢁ10 m), D (relative diffusion
coefficient of a complex and a water molecule, 2.24”
10ꢁ9 m2 sꢁ1), and r (Gd–H distance for an inner-sphere water
molecule, 3.1”10ꢁ10 m) and fixing the value of tM to that de-
termined previously from the 17O T2 value, we calculated the
electronic relaxation times at zero field (tso) and the tr
values of the complexes. The results are compiled in
Table 3. The relaxivities of GdL1 and GdL2 observed using
the NMRD technique are higher than that observed for
GdDTPA-BMA. This result can be ascribed to the higher
rotational correlation times of GdL1 and GdL2 since this pa-
Scheme 5. Structures of the DTPA-bisamide ligands mentioned in
Table 2.
We cannot exclude the possibility that the boronate func-
tion also catalyzes boronic prototropic exchange. However,
inclusion of additional terms in Equation (6) to account for
this catalytic effect did not lead to a better fit of the pH pro-
file of the relaxivity.
The parameters that influence the relaxivity can be evalu-
ated by measuring its field dependence (nuclear magnetic
Chem. Eur. J. 2004, 10, 5205 – 5217
www.chemeurj.org
ꢀ 2004 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
5211

J. A. Peters et al.
FULL PAPER
½BSŠ
Table 3. Parameters obtained from the best fit of the NMRD profiles
ð9Þ
Ka
¼
(258C) of GdL1 and GdL2 to the Solomon–Bloembergen–Morgan model.
½BŠ½SŠ
GdDOTA[a] and GdDTPAdata are included for comparison.
[38]
GdL1
GdL2
GdDTPA-BMA
GdDOTA
CtARS ¼ ½ARSŠ þ ½BARSŠ
CtS ¼ ½SŠ þ ½BSŠ
ð10Þ
ð11Þ
ð12Þ
tv [ps]
35ꢀ2
23ꢀ2
25ꢀ1
11ꢀ1
D2 [1019 sꢁ2
tso [ps]
]
2.3ꢀ0.1
105ꢀ7
214ꢀ11
4.5ꢀ0.1
81ꢀ12
186ꢀ22
4.1ꢀ0.2
81ꢀ5
1.6ꢀ0.1
473ꢀ52
77ꢀ4
CtB ¼ ½BŠ þ ½BSŠ þ ½BARSŠ
tr [ps]
66ꢀ11
[a] DOTA, tetraazacyclododecane.
I ¼ cARS Á IARS þ cBARS Á IBARS ¼ IARS þ cBARS Á ðIBARSꢁIARS
Þ
ð13Þ
[S], [B], [ARS], [BARS], and [BS] are the molar concen-
trations of the saccharide, the free boron functionalities
(both in the boronic and in the boronate form), free ARS,
the boronate ester formed with ARS, and the boronate ester
formed with the saccharide, respectively. CtARS, CtS, and CtB
are the total concentrations of ARS, the saccharide, and the
boronic acid groups, respectively. The molar fractions of
ARS and of its boronic acid ester are denoted as cARS and
cBARS, while IARS and IBARS are their specific emissions. K1
and Ka are the conditional association constants for associa-
tion of the boron function concerned with ARS or the sac-
charide, respectively. The formation of intra- and intermo-
lecular boronic esters with a boronic function/sugar ratio of
2:1 was neglected and the interactions of the two boronic
functions present in each complex were regarded as inde-
pendent. The concentration of the boronic acid was taken as
double that of the Gd3+ complex for the calculation. By fit-
ting the fluorescence data with Equations (8)–(13), we deter-
mined the formation constants of the adducts formed by
GdL1 and GdL2 with ARS to be 1958ꢀ205 and 12433ꢀ
2424 molꢁ1 L, respectively. Figure 4 shows the fluorescence
data for the binding of GdL1 and GdL2 with ARS in compe-
tition with the various sugar derivatives. We carried out a
least-squares fit of the data collected during the competition
experiments to the described model by using the above-
mentioned formation constants for association with ARS.
The results are listed in Table 4.
Both GdL1 and GdL2 interact considerably more strongly
with sialic acid than with the other saccharides studied.
Both complexes show good selectivity for sialic acid with re-
spect to two of the most common saccharides present in
glycan chains, methyl a-mannopyranoside and methyl b-gal-
actopyranoside. Surprisingly, the formation constants of
both complexes for association with sialic acid are also
about two to three times greater than those of their glucose
adducts. The a-furanose form of glucose is known to interact
relatively strongly with phenylboronic acid and its deriva-
tives since the cis-1,2-diol function in this anomeric form of
the sugar is highly preorganized for the formation of boro-
nate esters.[50–53] However, this anomeric form has an abun-
dance of only 0.14% at anomeric equilibrium. The more
abundant pyranose forms do not interact strongly with boro-
nates and, consequently, the overall stability constant of the
boronate ester is relatively low. These interactions may be
even weaker in vivo than in vitro since the anomeric equili-
brium of glucose is established very slowly and is rate deter-
mining for the formation of boronate esters of the furanose
form. In vitro measurements taken after a short equilibra-
rameter is the main factor governing the relaxivity under
these conditions.
The tso values are relatively low and, as stated above, are
typical for structures with low symmetry.[47] The reorienta-
tional correlation times (tr) of GdL1 and GdL2 are higher
than those of GdDOTAand GdDTPA, which is in accord-
ance with the higher molecular masses of GdL1 and GdL2.
Interaction of GdL1 and GdL2 with saccharides: The forma-
tion constants of the adducts of the two Gd3+ complexes
(GdL1 and GdL2) with Neu5Ac were compared with those
of the adducts formed with competing monosaccharides to
evaluate the in vivo selectivity of the ligands for sialic acid.
The interaction of Gd3+ complexes with macromolecular
substrates is often evaluated by means of the well-establish-
ed proton relaxation enhancement technique.[48,49] This ap-
proach exploits the increase in the observed water proton
longitudinal relaxation rate that occurs as a result of the in-
crease in the rotational correlation time upon formation of
the adduct. Unfortunately, the large tM values of GdL1 and
GdL2, combined with the small change in tr that occurs
upon interaction with monosaccharides hampered determi-
nation of the formation constants by this method. We there-
fore exploited an approach developed by Springsteen and
Wang,[26] which we adapted to suit our study.
The formation constants of the boronate esters of GdL1
and GdL2 and those of the saccharide adducts of these
esters were measured by means of competitive titrations at
pH 7.4. First, the formation constants of the esters formed
by the complexes upon treatment with Alizarine Red S
(ARS) were determined by measuring the change in fluores-
cence that occurred upon interaction. The association con-
stants of the esters of GdL1 and GdL2 with the sugar under
investigation were determined by titrating solutions of the
complex with both ARS and the sugar in question and mon-
itoring the change in fluorescence emission. In addition to
Neu5Ac, methyl b-d-galactoside, methyl a-d-mannoside,
and d(+)-glucose were studied for comparison. Methyl b-d-
galactoside and methyl a-d-mannoside are models of units
commonly present in glycan chains; glucose occurs at rela-
tively high concentrations in the blood. The systems can be
described by the equilibrium constants and mass balances
defined in Equations (8)–(13).
½BARSŠ
ð8Þ
K1
¼
½BŠ½ARSŠ
5212
ꢀ 2004 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chemeurj.org
Chem. Eur. J. 2004, 10, 5205 – 5217

MRI Contrast Agents
5205 – 5217
GdL1 interacts much less effectively than GdL2 or phenyl-
boronic acid with all the saccharides studied. Low preorga-
nization of the sensing moiety and/or steric hindrance
around the boronic group are possible causes of the weak
interactions observed for this complex.
The two boronate units present in each of the complexes
GdL1 and GdL2 may allow two-site binding of sialylated cell
surfaces. As a result, the binding of these complexes might
be considerably stronger than expected on the basis of the
association constants reported in Table 4.
Studies on cell interactions: We used a radioactive nuclide
and exploited the excellent sensitivity offered by radiomet-
ric methods to assess the capability of lanthanide complexes
of L1 and L2 to discriminate between cells presenting differ-
ent amounts of sialic acid on their surfaces. We investigated
the interaction of 153SmL1 and 153SmL2 with C6 rat glioma
cells, before and after treatment with neuraminidase. Like
most higher animal cells, C6 rat glioma cells have been re-
ported to have exposed sialic acid on their surfaces.[33]
In each experiment, one million cells were transferred to
each of the six wells of a multiwell plate. The cells in three
of the six wells were treated twice with neuraminidase ac-
cording to a published procedure[32] to remove most of the
sialic acid exposed on the cell surface. All cells were incu-
bated with the complex under investigation (2.6 pmol
153SmL1 or 153SmL2) for 30 or 60 min at 378C. Some experi-
ments were carried out at 08C to minimize the occurrence
of internalization phenomena. After removal of the test so-
lution, the cells were washed twice with phosphate-buffered
saline and detached from the plate by treatment with 1m
NaOH. The radioactivity of the cell suspension in NaOH
was measured to determine the amount of complex retained
by the cells.
Figure 4. Intensity of the emission at 616 nm upon excitation at 468 nm
of an aqueous solution of 1”10ꢁ5 m ARS, 1 mmol GdL1 (a) or GdL2 (b),
0.1m sodium phosphate monobasic buffer (pH 7.4), and various amounts
&
*
of N-acetylneuraminic acid ( ), d-glucose ( ), methyl a-d-mannoside
~
^
(
), or methyl b-d-galactoside ( ).
Amodified version of the procedure described above was
used to investigate whether internalization of the two com-
plexes (153SmL1 and 153SmL2) occurs at 378C. In these ex-
periments, the cells were incubated with the complex and
washed with phosphate-buffered saline, then treated with
glycine buffer at pH 2.8. At this pH value, the boronic func-
tional groups of both SmL1 and SmL2 are in their trigonal
forms (see above), which do not interact with vicinal diol
functions. Under these conditions, any SmL1 or SmL2 bound
to sialic acid end groups at the cell surface is removed. The
residual activity measured after detachment of the cells
from the plate affords an estimate of the amount of internal-
ized complex.
Table 4. Association constants (Ka) of some saccharides with GdL1 and
GdL2 in 0.10m phosphate buffer, pH 7.4. Each value shown is the average
of duplicate measurements. Standard deviations are given in parentheses.
Ka [molꢁ1 L][a]
GdL1
GdL2
N-acetylneuramic acid
d-glucose
Methyl a-d-mannoside
Methyl b-d-galactoside
3.30 (0.55)
1.70 (0.23)
0.19 (0.08)
0.43 (0.10)
50.43 (4.61)
15.19 (1.04)
-
[b]
0.96 (0.48)
[a] The boronic functions present in each complex were regarded as inde-
pendent (see the main text for details). [b] Not measured.
The results of these experiments are expressed as the dif-
ference between the residual activity of the cells not treated
with neuraminidase and that of the cells that were treated
(A, see Eq. (14)).
tion time can be expected to result in apparent formation
constants for association with glucose even lower than those
reported in Table 4. The relatively strong interaction of the
two complexes with sialic acid as compared to that with glu-
cose and the other two monosaccharides investigated is very
promising for the development of sialic acid sensors.
a1ꢁa2
A ¼
ꢃ 100
ð14Þ
a0
The interaction of GdL2 with sialic acid is considerably
stronger than that of phenylboronic acid with this sugar
(Ka =11.6mꢁ1). This observation reflects the synergetic
effect of the charged protonated amine functions of GdL2
on the binding of sialic acid. It is evident, however, that
The term a0 represents the total radioactivity of the so-
lution initially added to the cells, a1 is the final radioactivity
of cells not treated with neuraminidase, and a2 is the final
radioactivity of cells previously treated with neuraminidase.
Chem. Eur. J. 2004, 10, 5205 – 5217
www.chemeurj.org
ꢀ 2004 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
5213

J. A. Peters et al.
FULL PAPER
Each experiment was repeated at least twice and the mea-
sured A values listed in Table 5 are the average percent re-
sidual activities measured.
other competing saccharides. The interaction of the corre-
sponding 153Sm complex with C6 rat glioma cells appears to
be dependent on the concentration of sialic acid present on
the cell surface. L1 showed less favorable properties, al-
though the selectivity of its Gd3+ complex for sialic acid
with respect to competing saccharides is good. The reasons
for the less promising performance of L1 remain to be inves-
tigated further but the flexibility of the structure and steric
hindrance around the boronic function could play an impor-
tant role in reducing the affinity of its Ln3+ complexes for
sugars.
The relaxometric properties of the two complexes are
consistent with those of other Gd3+ complexes of DTPA-
bisamides. Since the relaxometric parameters of this class of
compounds are not optimal, a further improvement in the
performance of these systems might be achieved by conju-
gating the recognition moieties to ligands whose Gd3+ com-
plexes have a more favorable water exchange rate (20–
30 ns). Such complexes include phosphonate analogues of
DOTAand DTPA, as well as DOTAand DTPAstructures
with extended ethylene bridges. Further improvements in
the sensitivity of the complexes could be achieved by conju-
gating the targeting vectors and a number of Gd3+ chelates
to a high-molecular-weight carrier.
Table 5. Residual activity expressed according to Eq. (14). The data are
average values calculated from the results of at least two independent ex-
periments in which each measurement was carried out in triplicate. Stan-
dard deviations are given in parentheses.
Temp [8C]
A [%]
60 min
30 min
[a]
153SmL1
37
37
0
37
37
–
–
1.80 (0.95)
1.70 (0.89)
7.62 (1.15)
6.20 (0.30)
0.13 (0.08)
[a]
153SmL1 (glycine treatment)
153SmL2
2.02 (0.10)
3.27 (0.40)
0.22 (0.15)
153SmL2
153SmL2 (glycine treatment)
[a] Not determined.
The results for 153SmL2 point to a dominant effect of the
concentration of surface-bound sialic acid on the amount of
complex that is retained by the cells. For 153SmL1 the results
are less clear and, if any binding occurs, the relationship be-
tween the extent of binding and the sialic acid concentration
on the cell surface is much weaker than for 153SmL2. No sig-
nificant difference in residual activity was observed when
experiments were carried out with 153SmL2 at different tem-
peratures. The surface binding of this complex seems to be
unaffected by temperature.
Experimental Section
The results of the experiments carried out with glycine
treatment to investigate whether internalization processes
occur indicate that 153SmL2 is not internalized in any appre-
ciable quantity. The residual activity for 153SmL1 after treat-
ment with glycine suggests that a slow internalization proc-
ess exists. These results provide only a qualitative indication
of the influence of sialic acid on the binding of the two com-
plexes since the total quantity of sialic acid present on each
cell surface was not determined. However, our observations
concerning the surface binding of the two 153Sm3+ complexes
are in line with the relative values of the interaction con-
stants measured for the two Gd3+ complexes and sialic acid
(determined from the stability constants of the adducts con-
cerned; see above).
Materials: N-acetylneuraminic acid was purchased from Rose Sci. Ltd
and used without further purification. The cell culture medium, Dulbec-
coꢁs F12 minimal essential medium (DMEM-F12), was purchased from
Sigma and supplemented with 10% fetal bovine serum (FBS; Gibco), l-
glutamine (Sigma), and penicillin/streptomycin (Sigma). Earleꢁs balanced
salt solution (EBSS) medium was purchased from Sigma, as was neura-
minidase from vibrio cholerae. 153Sm oxide was produced at the Instituto
Tecnológico e Nuclear (ITN), Lisbon with a specific activity greater than
5 GBqmgꢁ1. Astock solution of 153SmCl3 was prepared by dissolving
153Sm2O3 in 0.1m HCl. All other reagent-grade chemicals were purchased
from commercial sources and used without further purification.
Physical methods: The NMR spectra were recorded on
a Varian
INOVA-300 spectrometer operating at 300, 75.5, and 96.2 MHz for 1H,
13C, and 11B, respectively, or on a Varian VXR-400S spectrometer operat-
ing at 400, 100.6, and 128.3 MHz for 1H, 13C, and 11B, respectively. 5-mm
sample tubes were used. For measurements in D2O, tert-butyl alcohol was
used as an internal standard, with the methyl signal calibrated at d=1.29
(1H) or 31.3 ppm (13C). 11B chemical shifts are reported with respect to
0.1m boric acid in D2O (taken as the external standard). The results can
be converted to the BF3·Et2O scale as follows: d(H3BO3 scale)=
d(BF3·Et2O scale)ꢁ18.7 ppm. Spectra were recorded at 258C unless
stated otherwise.
By assuming that the residual activity (6.2%) of the cells
treated with 2.6 pmol 153SmL2 at 378C corresponds to the
amount of this complex bound to sialic acid residues on the
cell surface, we estimated that 1.105 complex molecules are
bound by each cell.
The pH values of the samples were measured at ambient temperature by
using a Corning 125 pH meter with a calibrated microcombination probe
purchased from Aldrich. The pH values were adjusted by addition of
dilute solutions of NaOH and HCl.
Conclusions
We believe that recognition of sialic acid in living systems
by MRI contrast agents is of importance for the detection of
various malignancies. We designed and synthesized two new
ligands for lanthanide ions, L1 and L2, and characterized
their complexes. Gd3+ and 153Sm3+ complexes of L2 showed
very promising properties that may be useful in the develop-
ment of sialic acid sensors. GdL2 has a fairly high interaction
constant with Neu5Ac and is selective for the target over
The longitudinal water proton relaxation rates were measured on a
Stelar Spinmaster spectrometer (Mede) operating at 20 MHz. The stan-
dard inversion-recovery technique (16 experiments, two scans) was used.
The temperature was controlled with a Stelar VTC-91 air-flow heater
equipped with a copper thermocouple (uncertainty: ꢀ0.18C).
The proton 1/T1 NMRD profiles were recorded on a Stelar-Fast-Field-Cy-
cling relaxometer over a continuous range of magnetic field strengths
from 0.00024 to 0.28 T, which corresponds to a proton Larmor frequency
range of 0.01–12 MHz. The relaxometer was under complete computer
5214
ꢀ 2004 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chemeurj.org
Chem. Eur. J. 2004, 10, 5205 – 5217

MRI Contrast Agents
5205 – 5217
control and took measurements with an absolute uncertainty in 1/T1 of
ꢀ1%.
38.18 ppm (CH2). One of the aromatic carbon nuclei was not observed,
probably because of severe line broadening.
Variable-temperature 17O NMR measurements were recorded at 2.1 T
3,9-Bis{6-[(4,5-dihydroimidazol-2-yl)aminoethyl]-10-[2-(dihydroxybora-
nylphenyl)]-2-oxo-3,6,9-triazadecyl}-6-carboxymethyl-3,6,9-triazaundeca-
nedioic acid (L1): Asolution containing 4 (0.654 g, 1.47 mmol), DTPA-bis-
anhydride (0.262 g, 0.73 mmol), and zeolite KAin absolute ethanol
(16 mL) was stirred at room temperature for 6 h. The suspension was fil-
tered and the solvent was removed under reduced pressure. The residue
was taken up in water and lyophilized to give L1 as a pale yellow solid
(0.840 g, 95%). 11B NMR (300 MHz; D2O; pH 9.0): d=ꢁ11.4 ppm;
1H NMR (300 MHz; D2O; pH 6.4): d=7.32–7.43 (m, 2H; ArH), 7.13–
7.24 (m, 4H; ArH), 7.03–7.13 (m, 2H; ArH), 3.87 (s, 4H; ArCH2), 3.47
(s, 8H; CH2 imidaz), 2.44–3.39 ppm (brs, 42H; CH2N); 13C NMR
(300 MHz; D2O; pH 6.4): d=180.37 (CO), 175.33 (CO), 171.94 (CO),
161.33 (CN3), 145.80 (very broad; CB), 141.79 (Carom), 131.69 (CHarom),
129.68 (CHarom), 129.04 (CHarom), 125.62 (CHarom), 60.97 (CH2), 60.56
(CH2), 56.08 (CH2 central), 54.35 (CH2), 54.17 (CH2), 53.94 (CH2), 53.72
(CH2), 52.15 (CH2), 51.35 (CH2), 45.47 (CH2), 44.13 (CH2 imidaz), 41.63
(CH2), 38.38 ppm (CH2). ESI-MS (positive): m/z: 1056.7 [M+3]+.
with a Jeol EX-90 instrument equipped with a 5-mm probe. AD O exter-
2
nal lock was used. Experimental settings: spectral width, 10000 Hz; 908
pulse, 7 ms; acquisition time, 10 ms; 1000 scans; no sample spinning.
Aqueous solutions containing 2.6% 17O isotope (Yeda) were used. The
transverse relaxation rates were calculated from the signal widths at half
the maximum signal height.
Fluorescence emission spectra were recorded with a Spex/Jobin-Yvon
Fluorlog 3 fluorescence spectrometer (Instruments s.a.).
Awell counter (DPC-Gamma C) with a 12 Compaq DeskPro compatible
computer was used for activity counting in the interaction studies.
Molecular modeling was performed with the HyperChem software (ver-
sion 7.5) and the MM+ force field was used.
Di-tert-butyl-nitrilo-2,2’,2’’-triethanamine-N,N’-dicarboxylate (1): Com-
pound 1 was prepared by the procedure proposed by Hamdaoui et al.[34]
1H NMR (300 MHz; CDCl3): d=5.36 (brs, 2H; NH or NH2), 3.19 (m,
4H; CH2), 2.76 (t, 2H; CH2), 2.57 (4H, t; CH2), 2.53 (m, 2H; CH2), 2.06
(brs, 2H; NH or NH2), 1.47 ppm (s, 18H; CH3); 13C NMR (75.48 MHz;
CDCl3): d=160.61 (OCON), 83.42 (Me3CO), 61.19 (CH2N), 58.50
(CH2N), 44.04 (CH2N), 42.95 (CH2N), 32.73 ppm ((CH3)3O).
3,9-Bis(tert-butyl-2-oxo-3,6-diazahexyl-N-carboxylate)-6-carboxymethyl-
3,6,9-triazaundecanedioic acid (5): Compound 5 was prepared by follow-
ing the procedure described by Carvalho et al.[35] 1H NMR (300 MHz;
CDCl3): d=3.75 (s, 4H; CH2CO), 3.64 (s, 4H; CH2CO), 3.62 (s, 2H;
CH2CO), 3.16–3.32 (m, 12H; CH2N), 3.10 (t, 4H; CH2N), 1.30 ppm (s,
18H; CH3); 13C NMR (300 MHz; CDCl3): d=173.82 (CO), 173.73 (CO
central), 170.58 (CO), 159.82 (O2CN), 82.65 (CMe3), 58.42 (CH2CO),
58.19 (CH2CO), 56.23 (CH2CO central), 53.15 (CH2N), 52.97 (CH2N),
40.89 (2”CH2N), 29.38 ppm (CH3).
Di-tert-butyl-N’’-(4,5-dihydro-1H-imidazol-2-yl)-nitrilo-2,2’,2’’-triethan-
amine-N,N’-dicarboxylate (2): Asolution containing
23.57 mmol) and methylthioimidazoline hydroiodide
1
(8.167 g,
(5.812 g,
23.57 mmol) in EtOH (160 mL) was stirred at reflux (bath at 758C) for
5 h. The solvent was removed under reduced pressure and the resulting
red oil was used for the next step without further purification. 1H NMR
(300 MHz; CDCl3): d=3.78 (s, 4H; CH2 imidaz), 3.42 (m, 2H; CH2),
3.16 (m, 4H; CH2), 2.68 (t, 2H; CH2), 2.58 (t, 4H; CH2), 1.44 ppm (s,
18H; CH3); 13C NMR (75.48 MHz; CDCl3): d=160.07 (O2CN), 156.74
(CN3), 79.78 (OCMe3), 54.70 (CH2), 53.58 (CH2), 42.88 (CH2), 41.93
(CH2), 38.93 (CH2 imidaz), 28.53 ppm (CH3).
3,9-Bis(2-oxo-3,6-diazahexyl)-6-carboxymethyl-3,6,9-triazaundecanedioic
acid (6): Asolution of 5 (2.50 g, 3.69 mmol) in pure trifluoroacetic acid
(10 mL) was stirred for 12 h. The solvent was removed under reduced
pressure, followed by addition and then evaporation of CH2Cl2 (3”
10 mL) then ether (10 mL). The residue was taken up in water, filtered,
and lyophilized to give 6 (2.94 g, 97%) as a white solid. 1H NMR
(300 MHz; D2O; pH 1): d=4.14 (s, 4H; CH2CO), 4.11 (s, 4H; CH2CO),
3.80 (s, 2H; CH2CO), 3.62 (t, 4H; CH2N), 3.55 (t, 4H; CH2N), 3.30 (t,
4H; CH2N), 3.22 ppm (t, 4H; CH2N); 13C NMR (300 MHz; D2O; pH 1):
d=174.48 (CO), 171.94 (CO), 169.80 (CO), 57.95 (CH2N), 57.43 (CH2N),
55.58 (CH2N central), 54.15 (CH2N), 51.95 (CH2N), 40.53 (CH2N),
38.61 ppm (CH2N).
N-(4,5-Dihydro-1H-imidazol-2-yl)-nitrilo-2,2’,2’’-triethanamine (3): Com-
pound 2 was redissolved in EtOH (20 mL), then HCl (20 mL, 37%) was
added dropwise. This solution was stirred at RT for 1.5 h. The solvent
was removed under reduced pressure. MeOH (6”100 mL) was added
and then evaporated from the solution. The yellow foam obtained was
dissolved in a minimum amount of a methanol/water mixture and the
product was isolated by cation exchange chromatography on a DOWEX
50-W”8–200 (H+ form) column. An elution gradient from 20 to 50%
HCl (37% aq) in methanol was used. The solvent was removed under re-
duced pressure and the residue was taken up in water, then lyophilized.
The HCl salt of 3a (6.629 g, 87%) was obtained as a white powder.
1H NMR (400 MHz; D2O): d=3.72 (s, 4H; CH2 imidaz), 3.42 (t, 2H;
CH2), 3.08 (t, 4H; CH2), 2.80 (t, 4H; CH2), 2.74 ppm (t, 2H; CH2);
13C NMR (300 MHz; D2O): d=161.54 (CN3), 53.10 (CH2), 51.71 (CH2),
44.41 (CH2), 40.72 (CH2), 37.74 ppm (CH2).
3,9-Bis[3-(dihydroxoboranylphenyl)-2-oxo-3,6-diazaheptyl]-6-carboxy-
methyl-3,6,9-triazaundecanedioic acid (L2): Asolution of
6 (2.94 g,
3.59 mmol), 3-formylphenylboronic acid (1.38 g, 9.20 mmol), and triethyl-
amine (8.0 mL) in methanol (15 mL) was stirred at room temperature for
2 h. NaBH4 was added carefully (1.39 g, 36.7 mmol). The resulting so-
lution was stirred at RT for 12 h. The solvent was removed under re-
duced pressure and the residue was taken up in water (ca. 10 mL). The
resulting suspension was filtered and the desired product was recovered
from the solution by cation exchange chromatography on a DOWEX 50-
W”8–200 (H+ form) column. The column was washed with water, then
the product was eluted with an aqueous ammonia solution (8%). The
solvent was removed under reduced pressure. The residue was taken up
in water and lyophilized to give L2 (2.391 g, 90%) as a white powder.
N-(4,5-Dihydro-1H-imidazol-2-yl)-N’-(2-dihydroxyboranylphenyl)-nitrilo-
2,2’,2’’-triethanamine (4): Intermediate 3 (4.07 g, 12.56 mmol) was dis-
solved in a minimum amount of methanol/triethylamine (3:1 v/v, ca.
70 mL). Asolution of 2-formylphenylboronic acid (1.88 g, 12.56 mmol) in
the same solvent (320 mL) was added dropwise over 30 min. The mixture
was stirred at room temperature for 6 h. The solvent was removed under
reduced pressure and the residue was redissolved in methanol (100 mL).
NaBH4 (2.38 g, 62.9 mmol) was added carefully. The reaction mixture
was stirred overnight, then the solvent was removed under reduced pres-
sure. The residue was taken up in water (ca. 15 mL) and filtered. The
product was isolated by cation exchange chromatography on an Amber-
1
11B NMR (400 MHz; D2O; pH 11.3): d=ꢁ16.6 ppm; H NMR (300 MHz;
D2O; pH 11.3; T=808C): d=7.50–7.60 (brs, 4H; ArH), 7.52 (brs, 2H;
ArH), 7.18–7.25 (brs, 2H; ArH), 3.78 (s, 4H; ArCH2), 3.41 (t, 4H;
CH2N), 3.29 (s, 4H; CH2CO), 3.21 (s, 4H; CH2CO), 3.15 (s, 2H;
CH2CO), 2.81 (t, 4H; CH2N), 2.75 ppm (brs, 8H; CH2N); 13C NMR
(300 MHz; D2O; pH 11.3; T=808C): d=179.7 (CO), 175.7 (CO), 175.5
(CO), 138.6 (C), 132.3 (C), 131.1 (C), 128.2 (C), 127.0 (C), 126.7 (C), 60.1
(CH2), 59.5 (CH2), 59.1 (CH2), 53.5 (CH2), 52.9 (CH2), 47.8 (CH2), 39.4
(CH2), 30.7 ppm (CH2). ESI-MS (positive): m/z: 748.62 [M+3]+
+
lite CG50 (NH4 form) column with an elution gradient from 50 mm to
2m NH4OH in water. The fractions containing the desired product were
collected and lyophilized to give the carbonate salt of
4 (1.450 g,
Preparation of the Gd3+ complexes: Complexation was carried out by
addition of stoichiometric amounts of GdCl3 to aqueous solutions of the
ligands under weakly acidic conditions (5.5<pH<6.5) at room tempera-
ture. The formation of the complex was monitored by measuring the sol-
vent proton relaxation rate (1/T1). The small excess of free Gd3+ ions,
which yielded a noticeable increase in the observed relaxation rate, was
removed by centrifugation of the solution after adjustment to basic pH.
3.25 mmol, 26%) as a pale yellow fluffy solid. 1H NMR (300 MHz; D2O;
pH<2): d=7.71–7.76 (m, 1H; ArH), 7.38–7.51 (brs, 3H; ArH), 4.33 (s,
2H; ArCH2), 3.56 (s, 4H; CH2 imidaz), 3.32 (t, 2H; CH2), 3.27 (t, 2H;
CH2), 3.01–3.20 (brs, 6H; CH2), 2.92 ppm (t, 2H; CH2); 13C NMR
(300 MHz; D2O; pH 7): d=161.49 (CN3), 139.26 (Carom), 134.39 (Carom),
131.36 (Carom), 130.05 (Carom), 129.47 (Carom), 54.21 (CH2), 52.96 (CH2),
51.91 (CH2), 51.07 (CH2), 45.66 (CH2), 44.28 (CH2 imidaz), 41.37 (CH2),
Chem. Eur. J. 2004, 10, 5205 – 5217
www.chemeurj.org
ꢀ 2004 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
5215

J. A. Peters et al.
FULL PAPER
Competitive fluorescence assay: Amodified version of the procedure de-
veloped by Wang et al was used.[26] The association constants of GdL1
and GdL2 with Alizarine Red S (ARS) were first determined. Two solu-
tions, Aand B, were prepared. Solution Acontained ARS (ca. 1”10 ꢁ5 m)
and sodium phosphate monobasic buffer (0.1m) at pH 7.4. Solution B
was prepared by dissolving GdL1 or GdL2 in a portion of solution Ato
obtain a complex concentration of about 1 mm. The concentrations of
ARS and phosphate monobasic buffer remained the same. Solution B
was titrated into solution Ato produce solutions with a constant concen-
tration of ARS and a range of concentrations of GdL1 or GdL2. The in-
tensity of the emission at 616 nm was recorded. The excitation wave-
length was set at 468 nm.
The association constants for interaction of GdL1 or GdL2 with various
carbohydrates were measured in a second set of experiments. Aset of
solutions C was prepared by dissolving various amounts of the appropri-
ate saccharide in solution B (2.5 mL). At least seven different solutions C
were prepared for each saccharide, with a range of concentrations from 0
to 0.5m. Glucose solutions were allowed to stand at least 20 h to allow
them to reach anomeric equilibrium; the solutions C of the other sugars
were allowed to stand at least 1 h. Asialic acid solution was prepared by
dissolving the necessary amount of sialic acid in a minimum volume of
water and adjusting the pH to 7.4 by addition of 1m NaOH. This solution
was then lyophilized and the amount of salt present in the residue was
determined by recording the 1H NMR spectrum of a weighed sample in
the presence of a tert-butanol standard. Least-square analysis of the data
obtained for solutions C was performed with the program Scientist for
Windows by Micromath, version 2.0.
to Dr. L. Maat for his advice on nomenclature. C.F.G.C.G. acknowledges
support from the Foundation of Science and Technology, Portugal (pro-
ject no. POCTI/QUI/47005/2002) and Fondo Europeo de Desarrollo Re-
gional (FEDER).
[1] A. E. Merbach, É. Tóth, The Chemistryof Contrast Agents in Medi-
cal Magnetic Resonance Imaging, Wiley, Chichester, 2001.
[2] P. Caravan, J. J. Ellison, T. J. McMurry, R. B. Lauffer, Chem. Rev.
1999, 99, 2293.
[3] S. Aime, M. Fasano, E. Terreno, Chem. Soc. Rev. 1998, 27, 19.
[4] J. A. Peters, J. Huskens, D. J. Raber, Prog. Nucl. Magn. Reson. Spec-
trosc. 1996, 28, 283.
[5] V. Jacques, J. F. Desreux, Top. Curr. Chem. 2002, 221, 123.
[6] É. Tóth, L. Helm, A. E. Merbach, Top. Curr. Chem. 2002, 221, 61.
[7] M. Rudin, R. Weissleder, Nat. Rev. Drug Discovery 2003, 2, 123.
[8] R. J. Gillies, J. Cell. Biochem. 2002, Suppl. 39, 231.
[9] S. A. Wickline, G. M. Lanza, J. Cell. Biochem. 2002, Suppl. 39, 90.
[10] K. J. Potter, J. Cell. Biochem. 2002, Suppl. 39, 147.
[11] A. D. Nunn, K. E. Linder, M. F. Tweedle, J. Nucl. Med. 1997, 41,
155.
[12] R. Schauer in Carbohydrates in Chemistry and Biology, Vol. 3 (Eds.:
B. Ernst, G. W. Hart, P. Sinay), Wiley-VCH, Weinheim, 2000, p 227.
[13] R. Schauer, J. P. Kamerling, New Compre. Biochem. 1997, 29b, 243.
[14] R. Schauer, Glycoconjugate J. 2001, 17, 485.
[15] S. Kelm, R. Schauer, Int. Rev. Cytol. 1997, 175, 137.
[16] C. Traving, R. Schauer, Cell. Mol. Life Sci. 1998, 54, 1330.
[17] F. N. Lamari, N. K. Karamanos, J. Chromatogr. B, Anal. Technol.
Biomed. Life Sci. 2002, 781, 3.
[18] G. A. Lemieux, K. J. Yarema, C. L. Jacobs, C. R. Bertozzi, J. Am.
Chem. Soc. 1999, 121, 4278.
[19] A. P. Davis, R. S. Wareham, Angew. Chem. 1999, 111, 3160; Angew.
Chem. Int. Ed. 1999, 38, 2979.
[20] S. R. Haseley, Anal. Chim. Acta 2002, 457, 39.
[21] T. D. James, S. Shinkai, Top. Curr. Chem. 2002, 218, 159.
[22] J. Böeseken, A. van Rossem, Recl. Trav. Chim. Pays-Bas 1912, 30,
392.
[23] J. Böeseken, Chem. Ber. 1914, 46, 2612.
[24] H. G. Kuivila, A. H. Keough, E. J. Soboczenski, J. Org. Chem. 1954,
19, 780.
[25] J. O. Edwards, R. J. Sederstrom, J. Phys. Chem. 1961, 65, 862.
[26] G. Springsteen, B. Wang, Tetrahedron 2002, 58, 5291.
[27] G. Wulff, Pure Appl. Chem. 1982, 54, 2093.
[28] T. Burgemeister, R. Grobe-Einsler, R. Grotstollen, A. Mannschreck,
G. Wulff, Chem. Ber. 1981, 114, 3403.
[29] S. L. Wiskur, J. J. Lavigne, H. Ait-Haddou, V. Lynch, Y. H. Chiu,
J. W. Canary, E. V. Anslyn, Org. Lett. 2001, 3, 1311.
[30] K. Djanashvili, L. Frullano, J. A. Peters, unpublished results.
[31] C. L. Hannon, E. V. Anslyn, Bioorg. Chem. Front. 1993, 193.
[32] E. E. Simanek, G. J. McGarvey, J. A. Jablonowski, C.-H. Wong,
Chem. Rev. 1998, 98, 833.
[33] O. L. Berezovskaya, V. Mares, G. G. Skibo, J. Neurosci. Res. 1995,
42, 192.
[34] B. Hamdaoui, G. Dewynter, F. Capony, J.-L. Montero, C. Toiron, M.
Hnach, H. Rochefort, Bull. Soc. Chim. Fr. 1994, 131, 854.
Cell interaction studies: Asample of 153SmCl3 with a specific activity of
5 GBqmgꢁ1 was provided by the ITN in the form of a stock solution
(1.9”10ꢁ3 m) in 1n HCl.[54] 153SmL (L=L1 or L2) was prepared by adding
153SmCl3 (1 mCi) to a solution of the ligand in sodium acetate buffer
(0.4m, pH 5) to give a final ligand/153Sm mole ratio of 10:1. The radio-
chemical purity was determined by instant thin layer chromatography
(GelmanSciences) with ammonium acetate as the mobile phase. The
amount of bound metal averaged 98% for all complexes.
C6 rat glioma cells were grown in 75-mL bottles in DMEM-F12 supple-
mented with 10% FBS (Gibco), l-glutamine, and penicillin/streptomycin.
The day before the experiments were carried out, the cells were counted
and transferred to a six-well plate (106 cells per hole). DMEM-F12
(4 mL) was added to each well and the cells were incubated overnight at
378C under 5% CO2.
Half the cells (three wells of each plate) were treated twice with neura-
minidase according to the procedure described by Meyer et al.[55] The
medium was removed and the cells were washed twice with EBSS
medium (1.5 mL). Fresh EBSS medium (1.2 mL) was then added and the
cells were allowed to adjust to the medium for 1 h at 378C. SmL1 or
SmL2 (2.6 pmol, 3.333 KBq) was added to each well and the cells were in-
cubated for up to 1 h at 37 or 08C. After appropriate time periods (30
and 60 min), the experiment was stopped by removing the medium and
washing the cells twice with ice-cold phosphate-buffered saline (150 mL).
The cells were then treated with glycine buffer (150 mL, 0.05m glycine so-
lution, pH adjusted to 2.8 with 1n HCl) to distinguish between cell-sur-
face-bound (acid-releasable) and internalized (acid-resistant) radioligand.
We skipped this step (acid wash) in some of the experiments for compa-
rative purposes. Finally, the cells were treated with 1n NaOH (150 mL)
and incubated at 378C for 10 min to detach them from the plates. The ra-
dioactivity was measured with a g-counter. Each experiment was carried
out simultaneously in triplicate and was repeated at least twice.
[35] J. F. Carvalho, S. P. Crofts, S. M. Rocklage
(Salutar Inc.),
WO9110645, 1991.
[36] R. M. Smith, A. E. Martell, Critical StabilityConstants
Plenum, New York, 1975, p. 7.
,
Vol. 2,
[37] I. Bertini, C. Luchinat, Coord. Chem. Rev. 1996, 150, 1.
[38] D. H. Powell, O. M. Ni Dhubhghaill, D. Pubanz, L. Helm, Y. S. Leb-
edev, W. Schlaepfer, A. E. Merbach, J. Am. Chem. Soc. 1996, 118,
9333.
[39] S. Aime, E. Gianolio, A. Barge, D. Kostakis, I. C. Plakatouras, N.
Hadjiliadis, Eur. J. Inorg. Chem. 2003, 2045.
[40] T. J. Swift, R. E. Connick, J. Chem. Phys. 1964, 41, 2553.
[41] T. J. Swift, R. E. Connick, J. Chem. Phys. 1962, 37, 307.
[42] J. P. AndrØ, H. R. Maecke, É. Tóth, A. E. Merbach, J. Biol. Inorg.
Chem. 1999, 4, 341.
[43] S. Aime, M. Botta, M. Fasano, E. Terreno, Acc. Chem. Res. 1999, 32,
941.
Acknowledgement
This work was performed within the framework of EU COST Action
D18, “Lanthanide Chemistry for Diagnosis and Therapy.” The authors
thank Prof. R. N. Muller, Prof. L. Vander Elst, and Dr. S. Laurent (Uni-
versity of Mons-Hainaut, Mons, Belgium), Dr. C. Cabella and V. Mainero
(University of Turin, Turin, Italy) for help with the initial investigation of
the gadolinium complexes, and Dr. Maria Neves (Instituto Tecnológico e
Nuclear, Lisbon) for providing the 153Sm chloride. We are very grateful
5216
ꢀ 2004 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chemeurj.org
Chem. Eur. J. 2004, 10, 5205 – 5217

MRI Contrast Agents
5205 – 5217
[44] I. Solomon, Phys. Rev. 1955, 99, 559.
[45] N. Bloembergen, J. Chem. Phys. 1957, 27, 572.
[51] R. van den Berg, J. A. Peters, H. van Bekkum, Carbohydr. Res.
1994, 253, 1.
[52] J. C. Norrild, I. Søtofte, J. Chem. Soc. Perkin Trans. 2 2002, 303.
[53] J. C. Norrild, H. Eggert, J. Am. Chem. Soc. 1995, 117, 1479.
[54] F. C. Alves, P. Donato, A. D. Sherry, A. Zaheer, S. Zhang, A. J.
Lubag, M. E. Merritt, R. E. Lenkinski, J.V. Frangioni, M. Neves,
M. I. M. Prata, A. C. Santos, J. J. P. Lima, C. F. G. C. Geraldes,
Invest. Radiol. 2003, 38, 750.
[46] N. Bloembergen, L. O. Morgan, J. Chem. Phys. 1961, 34, 842.
[47] R. B. Clarkson, A. I. Smirnov, T. I. Smirnova, R. L. Belford in The
Chemistryof Contrast Agents in Medical Magnetic Resonance Imag-
ing (Eds.: A. E. Merbach, É. Tóth), Wiley, Chichester, 2001, p. 383.
[48] R. A. Dwek, Monographs on Physical Biochemistry: Nuclear Mag-
netic Resonance (N.M.R) in Biochemistry. Applications to Enzyme
Systems, Vol. 3, Oxford University Press, New York, 1973, p. 396.
[49] B. G. Jenkins, E. Armstrong, R. B. Lauffer, Magn. Reson. Med. 1991,
17, 164.
[55] D. C. Mayer, O. Kaneko, D. E. Hudson-Taylor, M. E. Reid, L. H.
Miller, Proc. Natl. Acad. Sci. USA 2001, 98, 5222.
Received: April 16, 2004
Published online: September 13, 2004
[50] G. de Wit, PhD Thesis, Delft University of Technology, 1979.
Chem. Eur. J. 2004, 10, 5205 – 5217
www.chemeurj.org
ꢀ 2004 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
5217