Relaxometric and modelling studies of the binding of a lipophilic Gd-AAZTA complex to fatted and defatted human serum albumin

Article abstract of DOI:10.1002/chem.200601277

A new lipophilic gadolinium chelate consisting of a long aliphatic chain bound to the AAZTA coordination cage (Gd-AAZTAC17) has been synthesised. It possesses two coordinated water molecules (q = 2) in fast exchange with the solvent (τ_M²⁹⁸ = 67 ns), which yields a relaxivity of 10.2 mM^-1 s^-1. At concentrations greater than 0.1 mM, it forms micelles (average diameter 5.5 nm) characterised by a relaxivity of approximately 30 mM^-1 s^-1 at 20 MHz and 298 K. The latter value appears to be "quenched" by magnetic interactions among the Gd^III ions on the surface of the micelle that cause a decrease in the electronic relaxation time. A relaxivity of 41 mM^-1 s^-1 was recorded for this micellar system when 98% of the Gd^III ions were replaced by diamagnetic Y^III. Gd-AAZTAC17 exhibits a better affinity for fatted human serum albumin (HSA) than for defatted HSA, whereas the relaxivities of the supramolecular adducts are reversed. The relaxivity shown by Gd-AAZTAC17/defatted HSA (r₁^b(20 MHz, 298 K) = 84 mM ^-1 s^-1) is by far the highest relaxivity reported so far for non-covalent paramagnetic adducts with slow-moving substrates. As shown by molecular docking calculations, the gadolinium complex enters a hydrophobic pocket present in fatted HSA more extensively than the corresponding adduct with defatted HSA. Interestingly, no marked difference was observed in either the relaxation enhancement or the binding affinity between fatted and defatted HSA when the binding titrations were carried out at a Gd-AAZTAC17 concentration higher than its critical micellar concentration (cmc). This behaviour has been attributed to the formation of an association between the negatively charged micelle of the lipophilic metal complexes and the positive residues on the surface of the protein.

Full text of DOI:10.1002/chem.200601277

FULL PAPER
DOI: 10.1002/chem.200601277
Relaxometric and Modelling Studies of the Binding of a Lipophilic
Gd-AAZTA Complex to Fatted and Defatted Human Serum Albumin
Eliana Gianolio,[a] Giovanni B. Giovenzana,[b] Dario Longo,[a] Irene Longo,[a]
Ivan Menegotto,[b] and Silvio Aime*[a]
Abstract: A new lipophilic gadolinium
chelate consisting of a long aliphatic
chain bound to the AAZTA coordina-
tion cage (Gd-AAZTAC17) has been
synthesised. It possesses two coordinat-
ed water molecules (q=2) in fast ex-
change with the solvent (t2M98 =67 ns),
41 mmꢀ1 sꢀ1 was recorded for this micel-
lar system when 98% of the GdIII ions
were replaced by diamagnetic YIII. Gd-
AAZTAC17 exhibits a better affinity
for fatted human serum albumin
(HSA) than for defatted HSA, whereas
the relaxivities of the supramolecular
adducts are reversed. The relaxivity
shown by Gd-AAZTAC17/defatted
HSA (rb1(20 MHz, 298 K)=84 mmꢀ1 sꢀ1)
is by far the highest relaxivity reported
so far for non-covalent paramagnetic
adducts with slow-moving substrates.
As shown by molecular docking calcu-
lations, the gadolinium complex enters
a hydrophobic pocket present in fatted
HSA more extensively than the corre-
sponding adduct with defatted HSA.
Interestingly, no marked difference was
observed in either the relaxation en-
hancement or the binding affinity be-
tween fatted and defatted HSA when
the binding titrations were carried out
which
yields
a
relaxivity
of
10.2 mmꢀ1 sꢀ1. At concentrations great-
er than 0.1 mm, it forms micelles (aver-
age diameter 5.5 nm) characterised by
at
a Gd-AAZTAC17 concentration
a
relaxivity
of
approximately
higher than its critical micellar concen-
tration (cmc). This behaviour has been
attributed to the formation of an asso-
ciation between the negatively charged
micelle of the lipophilic metal com-
plexes and the positive residues on the
surface of the protein.
3 0 mmꢀ1 sꢀ1 at 20 MHz and 298 K. The
latter value appears to be “quenched”
by magnetic interactions among the
GdIII ions on the surface of the micelle
that cause a decrease in the electronic
Keywords: albumin
·
contrast
agents · gadolinium · magnetic reso-
nance imaging · micelles
relaxation time.
A
relaxivity of
Introduction
CAs and it is expected that this percentage will further in-
crease with the availability of more sensitive and specific
agents.[1–4] A MRI CA is not directly visualised in the image,
only its effects on water proton relaxation times are ob-
served. The increased relaxation rates allow the attainment
of an intense signal in a short time and a better signal-to-
noise ratio by the acquisition of a higher number of meas-
urements.
It is now well established that the information content of a
magnetic resonance imaging (MRI) experiment can be sig-
nificantly enhanced by the use of a suitable contrast agent
(CA). Nowadays about 35% of MRI scans make use of
[a] Dr. E. Gianolio, Dr. D. Longo, Dr. I. Longo, Prof. S. Aime
Dipartimento di Chimica I.F.M. e
As unpaired electrons display a remarkable ability to
reduce T1 and T2, the search for efficient CAs has focussed
on paramagnetic metal complexes. The metal of choice has
been the GdIII ion for its high paramagnetism (seven un-
paired electrons) and for its favourable properties in terms
of electronic relaxation.[5] This ion forms very stable com-
plexes with polyamminocarboxylate ligands and Gd-DTPA
(Magnevist) was the first MRI CA approved for clinical use.
Since then, other GdIII-based CAs similar to Magnevist have
been introduced into clinical practice. They have very simi-
lar pharmacokinetic properties because they are distributed
in the extracellular fluid and are eliminated via glomerular
Centro di eccellenza per l’Imaging Molecolare
Università degli Studi di Torino
Via P. Giuria 7, 10125 Torino (Italy)
Fax : (+39)116-707-855
E-mail: silvio.aime@unito.it
[b] Prof. G. B. Giovenzana, Dr. I. Menegotto
Dipartimento di Scienze Chimiche Alimentari, Farmaceutiche e
Farmacologiche
Università del Piemonte Orientale “A. Avogadro”
Via Bovio 6, 28100 Novara, (Italy)
Supporting information for this article is available on the WWW
under http://www.chemeurj.org or from the author.
Chem. Eur. J. 2007, 13, 5785 – 5797
ꢀ 2007 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
5785

filtration. The relative ability of a paramagnetic metal com-
plex to act as a MRI CA is expressed by its relaxivity, that
is, the property that is a measure of the relaxation enhance-
ment of water protons in solution containing the paramag-
netic agent at a 1 mm concentration.[6] Much work has been
carried out in the last two decades to improve the relaxivity
of GdIII-based agents. It was recognised early on that high
relaxivities can be attained for slow-moving systems endow-
ed with fast exchange rates of the coordinated water mole-
cule(s) and suitably long electronic relaxation rates of the
unpaired electrons on the metal ion.[3,4,7] On this basis sever-
al macromolecular systems have been designed in which the
GdIII chelates are either covalently or non-covalently bound
to high molecular weight substrates. In particular much at-
tention has been devoted to supramolecular adducts formed
by suitably functionalised GdIII chelates and human serum
albumin (HSA).[8] Although the theory of paramagnetic re-
laxation foresees for the latter systems r1 values of the order
of 100 mmꢀ1 sꢀ1 (at the observation frequency of 20 MHz),
the relaxivities of the numerous reported systems have
never reached such high values. The highest reported r1
value (65.8 mmꢀ1 sꢀ1)[9] is relative to the adduct of HSA with
a Gd-TTDA-BOM derivative. Often low r1 values have
been ascribed to an insufficiently fast exchange of the coor-
dinated water molecules[10] and this drawback has generated
a number of studies aimed at the understanding of the rela-
tionships between the structure of the chelate and the life-
time of the coordinated water. In this context, it has been
suggested that the problems in attaining the theoretically ex-
pected relaxivities might also be due to the inner rotation of
the water molecule around its axis of coordination to the
metal ion.[11,12] Such local motion superimposed on the over-
all molecular tumbling quenches the relaxation enhance-
ment expected for slow-moving macromolecular systems. If
this is the case, the high relaxivity target has to be pursued
by exploring other routes. For instance, by increasing the
number of water molecules coordinated to the GdIII ion as
relaxivity increases with the hydration state. An increase in
the number of coordinated water molecules can be obtained
by reducing the denticity of the ligand and this often occurs
at the expense of the overall thermodynamic stability which
is one of the primary requisites of a GdIII complex for it to
be considered as a potential CA for in vivo investigations.
In this context the first system to be investigated was Gd-
DO3A, but it was soon realised that, in a biological fluid, its
coordinated water molecules are replaced by endogenous
ions like lactate and carbonate.[13,14] A related system is rep-
resented by Gd-PCTA whose coordinated water molecules
are not so easily replaced by endogenous ligands.[15–17] How-
ever, its further functionalisation for binding to a macromo-
lecular substrate proved to be rather difficult. In addition to
polyaminocarboxylate ligands, there is the interesting class
of HOPO ligands developed by Raymond and co-work-
ers.[18–20] Their complexes with GdIII are outstandingly stable
and their potential for MRI applications appears very good.
Recently we reported a new heptadentate ligand, namely
AAZTA,[21] characterised by several properties favourable
for use as a MRI CA. Herein we report on our studies on a
lipophilic Gd-AAZTA-based system aimed at assessing the
relaxation enhancement capabilities of slow-moving para-
magnetic adducts formed by its self-assembly and its binding
to HSA.
Results and Discussion
Synthesis: The synthesis of AAZTAC17 is analogous to that
reported for the parent AAZTA ligand.[21] Therefore it is
simple and relies on readily available and cheap chemicals.
First 1-nitrooctadecane was synthesised by oxidation of oc-
tadecanamine with m-chloroperoxybenzoic acid (mCPBA)
in refluxing 1,2-dichloroethane (Scheme 1). The aliphatic ni-
Scheme 1. Scheme of the synthesis of AAZTAC17 ligand
troalkane was then employed in the key step involving the
formation of the seven-membered ring through the high-per-
forming double nitro-Mannich reaction with N,N’-dibenzyl-
ethylenediamine and paraformaldehyde. Catalytic transfer
hydrogenation of the nitro group and combined hydrogenol-
ysis of the N-benzyl groups was achieved with ammonium
formate and Pd/C. The triamine was then alkylated by reac-
tion with tert-butyl bromoacetate. Finally the tert-butyl
esters were removed in neat TFA to give the desired ligand.
As for the parent AAZTA ligand, the gadolinium com-
plex of AAZTAC17 is prepared by simply stirring stoichio-
metric amounts of the ligand and GdCl3 at neutral pH at
ambient temperature.
5786
www.chemeurj.org
ꢀ 2007 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Chem. Eur. J. 2007, 13, 5785 – 5797

Binding of a Gd-AAZTA Complex to HSA
FULL PAPER
Relaxometric characterisation: The presence of the aliphatic
chain is responsible for the concentration effects on the re-
laxation enhancement properties of Gd-AAZTAC17. In
fact, functionalisation of the AAZTA ligand with the C17
moiety induces the formation of micelles already at submilli-
molar concentrations. The critical micellar concentration
(cmc) of the system was conveniently determined by meas-
1
uring the H relaxation rates as a function of GdIII complex
concentration (Figure 1).
Figure 2. Temperature dependence of the transverse 17O water relaxation
*
rate (R2p) for a 7 mm solution of Gd-AAZTAC17 ( ) and Gd-AAZTA
*
(
) at 2.1 T and neutral pH.
At the field used for the relaxometric experiments
(0.47 T), the 1H relaxivity of GdIII complexes is given by
Equation (2), where K is a constant related to the strength
of the dipolar interaction between the protons and the elec-
tronic magnetic moment and tR is the correlation time asso-
ꢀ
ciated with the reorientation of the Gd H magnetic vector.
Figure 1. Plot of the variation of the observed longitudinal proton relaxa-
tion rate as a function of the concentration of Gd-AAZTAC17 at 0.47 T
and 258C.
1
TH1M
r1p
/
¼ KfðtRÞ
ð2Þ
The effect of aggregation may be estimated by analysing
the profile of the relaxivity data as a function of the applied
field strength. In fact from a quantitative analysis of the
NMRD (nuclear magnetic relaxation dispersion)[22] profiles
it is possible to determine the reorientational correlation
time (tR) which is related to the size of the investigated
system. The NMRD profiles of Gd-AAZTAC17 at 25 and
378C are reported in Figure 3. For the micellar system, data
were analysed using the Solomon–Bloembergen–Morgan
model modified according to the Lipari–Szabo approach
(see Supporting Information). This approach distinguishes
between two independent motions: a rapid local motion
governed by a correlation time defined as tl and a slower
In general, titration of a ligand solution with GdIII ions
leads to the formation of a paramagnetic complex and a
straight line for the relaxivity versus GdIII complex concen-
tration plot is obtained [Eq. (1)] whose slope corresponds to
the relaxivity (r1p) of the complex, where R1w is the diamag-
netic contribution of pure water (0.38 sꢀ1).[22]
Rlobs ¼ ½Gd-LŠr1p þ R1w
ð1Þ
In the case of a self-assembling system such as Gd-
AAZTAC17, sharp variation of the linear slope is observed
when the system passes from the monomeric state to micel-
lar aggregates. From the data reported in Figure 1, a cmc
value of 0.108 mm was determined.
At 20 MHz and 298 K, the relaxivity of Gd-AAZTAC17
below the cmc is 10.2 mmꢀ1 sꢀ1, whereas the relaxivity of the
self-assembled complex above the cmc is 30.0 mmꢀ1 sꢀ1. The
relaxivity of the non-associated form is around 30% higher
than that of the parent Gd-AAZTA which is fully consistent
with the increase in molecular weight caused by the intro-
duction of the long aliphatic chain.
From an analysis of the temperature dependence of the
transverse relaxation rate of the metal-bound 17O water res-
onance[23,24] (Figure 2) we obtained a t2M98 value of 67 ns for
Gd-AAZTAC17 above the micellar concentration, a value
that is even smaller than that reported for the parent com-
plex Gd-AAZTA (90 ns)[21] and so is in the optimal range
for the attainment of high relaxivities in the presence of
long molecular reorientational times.
Figure 3. 1/T1 NMRD profiles of a 1 mm solution of Gd-AAZTAC17 mi-
&
celles measured at 25 ( ) and 378C (&), of the Gd-AAZTAC17 mono-
*
*
mer measured at 25 ( ) and 378C ( ) and of Gd-Y-AAZTAC17 micelles
~
at 258C ( ). The solid curves through the data points were calculated
using the parameters reported in Table 2.
Chem. Eur. J. 2007, 13, 5785 – 5797
ꢀ 2007 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chemeurj.org
5787

S. Aime et al.
Table 1. Principal relaxometric parameters for the different Gd-AAZTAC17-containing systems determined
by NMRD and 17O-R2p-VT analysis.
towards the interaction with bi-
carbonate and hydroxide anions
present in solution.
r1p
D2
[sꢀ2
tv
tl
tg
S
tM
q
(20 MHz)
G
]
[ps]
[ps]
[ps]
[ns]
U
Assessment of micellar size and
aggregation number (Nagg): Size
measurements were performed
using a dynamic light scattering
instrument (DLS) (also known
as photon correlation spectros-
copy), which measures Browni-
an motion in solution and re-
lates them to the sizes of the
particles. The method involves
illuminating the particles with a
laser and analysing the intensity
fluctuations in the scattered
light. Particles suspended in a
liquid are never stationary as
Gd-AAZTAC17 30.0
micelles
258C
Gd-AAZTAC17 25.0
micelles
378C
1.43ꢁ0.28”1019 43ꢁ3.1 295ꢁ60 2540ꢁ130 0.40ꢁ0.15 67ꢁ19
2[a]
1.83ꢁ0.13”1019 37ꢁ1.2 222ꢁ30 2338ꢁ255 0.40ꢁ0.14 29ꢁ7.1 2[a]
Gd-YAAZ-
TAC17
41.4
9.53ꢁ3.5”1018 42ꢁ5.3474 ꢁ75 2530ꢁ350 0.48ꢁ0.16
67[a]
2[a]
micelles
258C
Gd-AAZTAC17 10.2
monomer
258C
1.51ꢁ0.1”1019 35ꢁ1.3142
3.87ꢁ0.14”1019 23ꢁ0.75
ꢁ1.6
–
–
–
–
2[a]
Gd-AAZTAC17
monomer
378C
7.35
100ꢁ1.15
2[a]
[a] These parameters were fixed during the fitting procedure.
global motion with a correlation time defined as tg. Data re-
sulting from the fitting of the NMRD profiles are reported
in Table 1 and are in accordance with other similar micellar
systems.[25,26] Clearly micellisation of the C17 complex gives
rise to a system with a slower molecular tumbling relative to
the monomeric complex, but the resulting relaxivity appears
still lower than the expected values as internal motions are
faster than the overall tumbling of the micellar system. Re-
cently Merbach and co-workers demonstrated that interac-
tions between nearby paramagnetic centres in micellar sys-
tems increase the transverse electronic relaxation of the
electron spins of GdIII and, therefore, reduce the attainable
water proton relaxivity.[27] This drawback may be removed
by diluting the GdIII ions with diamagnetic YIII ions in order
to increase the average distance between neighbouring GdIII
ions. The NMRD profile of mixed micelles formed with 2%
GdIII and 98% YIII at the same temperature and pH is re-
Figure 4. pH dependence of the proton relaxivity (r1p) of Gd-AAZTAC17
micelles measured at 0.47 T and 258C.
they are constantly moving as a result of Brownian motion.
The relationship between the size of a particle and its speed
due to Brownian motion is defined by the Stokes–Einstein
equation.
~
ported in Figure 3( ). An enhancement in the relaxivity of
around 40% over all the frequency range was observed.
The effect of replacing gadolinium with yttrium is two-
fold, namely, 1) it lengthens the electronic relaxation time
(and this mainly affects r1p values at low magnetic fields)
and 2) it gives rise to a change in the reorientational time
that causes an increase in r1p at high magnetic fields. The
latter effect appears to be associated with an anisotropic
tumbling of the yttrium/gadolinium mixed micelles with the
gadolinium-containing monomers aligned along the main ro-
tational axis.
The relaxivity of Gd-AAZTA is constant in the pH range
4–12 (Figure 4), suggesting an overall stability of the com-
plex in this pH range. Progressive protonation of the carbox-
ylic groups at pH values of less than 4 causes the neutralisa-
tion of the system and the consequent formation of a precip-
itate. At basic pH, in contrast to what is reported for other
q=2 gadolinium complexes such as Gd-DO3A,[11] the relax-
ivity remains constant, suggesting nearly complete inertness
A correlator, within the instrument, measures the degree
of similarity between two signals recorded in the same posi-
tion over a period of time. As particles move within the
sample, therefore, the correlation decreases with time. A
perfect correlation is reported as 1 and no correlation is re-
ported as 0. Correlation functions for large and small parti-
cles will therefore be different as the rate of decay is much
faster for small particles than it is for large ones. After the
correlation function has been measured this information can
then be used to calculate the size distribution.
Measurements were taken at room temperature on solu-
tions of Gd-AAZTAC17 at different concentrations above
the cmc. The size distribution diagram showed that one
main species is present that has a mean diameter of (5.5ꢁ
0.15) nm (see the Supporting Information).
As static light scattering (SLS) is a valid method[28–31] for
determining the apparent weight-average molar mass of a
5788
www.chemeurj.org
ꢀ 2007 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Chem. Eur. J. 2007, 13, 5785 – 5797

Binding of a Gd-AAZTA Complex to HSA
FULL PAPER
solute (Mapp), the aggregation number (Nagg
AAZTAC17 micelles has been calculated by dividing Mapp
by the molecular weight of Gd-AAZTAC17.
)
of Gd-
Fitting of experimental data (see the Supporting Informa-
tion) gave an apparent molecular weight of (70.9ꢁ
1.203) kDa. From the ratio of this value to the molecular
weight of Gd-AAZTAC17, an Nagg (aggregation number)
value of 95 was determined.
The Rayleigh expression describing the intensity of light
scattered by a particle in solution is given by Equation (3),
where K is an optical constant defined in Equation (4), C is
the particle concentration, Rq is the Rayleigh ratio of scat-
tered to incident light intensity, M is the average molecular
weight, A2 is the second virial coefficient, 1/Pq is an angle-
dependent term defined in Equation (5), l0 is the wave-
length of the incident light in a vacuum, NA is Avogadroꢁs
number, n0 is the solvent refractive index, dn/dc is the sol-
vent and analyte-dependent refractive index increment, Rg
is the radius of gyration and q is the scattering angle.
Binding to HSA: Owing to the possible angiographic MRI
applications for this system, the interaction with human
serum albumin has been investigated in detail both in the
case of the monomeric gadolinium complex (below the cmc)
and in the case of the aggregated micellar system (above the
cmc).
The binding of Gd-AAZTAC17 to HSA below the critical
micellar concentration was conveniently assessed by measur-
ing the water proton relaxation rates of solutions containing
the paramagnetic complex and increasing concentrations of
the serum protein (PRE method).[10] In addition to yielding
an estimate of the binding strength (nKA, where n is the
number of binding sites) these measurements also provide a
direct assessment of the relaxivity of the macromolecular
adduct (rb1).
ꢀ
ꢁ
KC
Rq
1
M
1
Pq
¼
þ 2A2C
ð3Þ
ð4Þ
ð5Þ
ꢀ
ꢁ
2
2p2
l40NA
dn
K ¼
n
0 dc
ꢀ
ꢁ
16p2n20R2g
3l20
Analysis of the relaxometric data obtained from the titra-
tion of a 20 mm aqueous solution of the GdIII complex with
fatted and defatted HSA at pH 7.4 and 298 K (Figure 5) al-
1
Pq
q
¼ 1 þ
sin2
2
The Rayleigh ratio (Rq) term in Equation (3) is the ratio
of the scattered to incident light intensity. As the intensity
of incident light interacting with a molecule is difficult to
measure, the standard approach is to measure the scattering
intensity of the analyte relative to that of a well-described
standard with a known Rayleigh ratio. A common standard
used in light scattering is toluene. The expression used to
calculate the sample Rayleigh ratio from a toluene standard
is given in Equation (6), where IA is the residual scattering
intensity of the analyte (sample intensity-solvent intensity),
IT is the toluene scattering intensity, n0 is the solvent refrac-
tive index, nT is the toluene refractive index and RT is the
toluene Rayleigh ratio.
IAn2
Figure 5. Plot of the variation of the observed longitudinal water proton
relaxation rate of a 20 mM solution of Gd-AAZTAC17 as a function of
Rq ¼
0 RT
ð6Þ
ITn2T
*
*
fatted ( ) and defatted ( ) HSA concentration (0.47 T, 258C).
The 1/Pq term in Equation (5) embodies the angular de-
pendence of the sample scattering intensity. The angular de-
pendence arises from different points within a particle or
molecule. This phenomenon is known as Mie scattering.
When the particles in solution are much smaller than the
wavelength of the incident light (for particles with hydrody-
namic diameter <80 nm the molecular weight errors are
<5%),[32] Mie scattering is minimised and 1/Pq reduces to 1.
So, for pure Rayleigh scattering, Equation (3) can be re-
duced to Equation (7) such that a plot of KC/Rq versus C
(Debye plot) is expected to be linear with an intercept
equivalent to 1/M.
lowed us to determine the nKA and rb1 values reported in
Table 2. The binding affinity increases significantly on going
from defatted to fatted HSA, whereas the relaxivity of the
paramagnetic macromolecular adduct is markedly higher for
defatted HSA. Binding of this lipophilic gadolinium com-
plex to defatted HSA leads to the highest relaxivity ever re-
ported for a stable gadolinium chelate bound to a slow-
moving substrate. The value of 84 mmꢀ1 sꢀ1 (at 20 MHz) is
30% larger than the most efficient system reported until
now.
Although it has been possible to identify the domain of
the interaction of Gd-AAZTAC17 with HSA on the basis of
competition assays with palmitic acid, warfarin and ibupro-
KC
Rq
1
M
ð7Þ
¼
þ 2A2C
Chem. Eur. J. 2007, 13, 5785 – 5797
ꢀ 2007 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chemeurj.org
5789

S. Aime et al.
Table 2. Binding parameters for Gd-AAZTAC17 (as a monomer and as
micelles) and Gd-Y-AAZTAC17 systems determined using the PRE
method at 258C in phosphate buffer solutions.
Further insights into the identification of the binding site
were obtained by carrying out a competitive assay with war-
farin, a well-established ligand for binding to HSA Sudlowꢁs
site I.[39] As shown in Figure 7, warfarin removes the gadoli-
Defatted
HSA
Fatted
HSA
nKA
2.3ꢁ0.7
84ꢁ3.3
7.1ꢁ1.8
63ꢁ1.3
1.7ꢁ0.6
35.5ꢁ0.21
–
Gd-AAZTAC17
(under cmc)
[104 mꢀ1
]
rb1 [mmꢀ1 sꢀ1
]
]
]
nKA
1.5ꢁ0.4
Gd-AAZTAC17 (above cmc) [105 mꢀ1
]
rb1 [mmꢀ1 sꢀ1
36.8ꢁ0.34
1.1ꢁ0.26
42.3ꢁ0.45
nKA
Gd/Y-AAZTAC17 (above
cmc)
[105 mꢀ1
]
rb1 [mmꢀ1 sꢀ1
–
fen, access to details of the binding mode has only become
possible thanks to molecular docking studies (see below).
The binding of fatty acids to serum albumin has been
studied for over 40 years, but a complete understanding of
the determinants of these interactions has not yet been
reached. Although it is now well established that the protein
has multiple fatty acid binding sites of different affinity,[33–35]
the precise number of binding sites is not known. The cur-
rent consensus seems to be that there are two or three high
affinity sites and at least three further sites of lower affini-
ty.[34,35] In particular, two of these lower-affinity binding sites
(named as sites 6 and 7) seem to be the favourite ones for
the binding of the complex under scrutiny. These two sites
are located on albumin in the region of subdomain IIA[33–38]
(corresponding to Sudlowꢁs site I).
Upon addition of three equivalents of palmitic acid to the
defatted HSA the resulting relaxivity of the supramolecular
adduct with Gd-AAZTAC17 is the same as that found for
the fatted HSA-containing system. Furthermore, addition of
palmitic acid to the solution containing the fatted HSA/Gd-
AAZTAC17 adduct leads to a progressive decrease in the
observed relaxivity (Figure 6).
Figure 7. Plot of the variation of the observed longitudinal water proton
*
*
)
relaxation rate of Gd-AAZTAC17 bound to fatted ( ) and defatted (
HSA as a function of increasing concentration of warfarin (0.47 T, 258C).
nium complex with either fatted and defatted HSA although
the r1 decrease is slower in the case of fatted HSA, as was
expected on the basis of the higher affinity shown by Gd-
AAZTAC17 for this form. The addition of the same
amounts of ibuprofen, a classical competitor for Sudlowꢁs
site II, to the adducts of Gd-AAZTAC17 with fatted and de-
fatted HSA showed no competition effect.
Furthermore, it is known that, at low concentrations,[40,41]
the presence of long-chain fatty acids such as palmitic acid
(C16) actually enhances the binding of warfarin and other
site I binding molecules (i.e., bilirubin and triiodobenzoic
acid); at high molar ratios long-chain fatty acids start to dis-
place the drug.[42] The Ka values obtained for the compound
under study towards fatted and defatted HSA suggest that
similar behaviour may also occur in the binding of Gd-
AAZTAC17; in fact, in the case of fatted HSA (2–3equiva-
lents of fatty acid are normally bound to human albumin in
serum), the observed affinity is higher than with defatted
HSA.
Allosteric effects related to the presence of fatty acids on
the protein are due to the dramatic conformational changes
they induce in the protein structure;[43,44] the rotations of the
three domains of the protein cause the opening of the cen-
tral cervix with the width of the molecule increasing from 80
to 90 Š. These conformational changes are responsible for
the difference in the binding affinity of Gd-AAZTAC17
with fatted and defatted HSA.
The NMRD profiles of the adducts of Gd-AAZTAC17
with fatted and defatted HSA have been recorded in the
range of 0.01–70 MHz at 258C in phosphate buffer
(Figure 7).
The experimental data reported in Figure 8 can be ana-
lysed only qualitatively as the investigated system appears
to be too complex for quantitative fitting. Both profiles dis-
play a high relaxivity peak centred at around 30 MHz, typi-
Figure 6. Plot of the variation of the observed longitudinal water proton
&
relaxation rate of Gd-AAZTAC17 bound to fatted ( ) and defatted (<)
HSA as a function of increasing concentration of palmitic Acid (0.47 T,
258C).
5790
www.chemeurj.org
ꢀ 2007 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Chem. Eur. J. 2007, 13, 5785 – 5797

Binding of a Gd-AAZTA Complex to HSA
FULL PAPER
using a model that provides contributions from inner,
second and outer coordination spheres. As shown in the
graphical separation of these three contributions, the de-
crease in relaxivity in the first part of the curve (low temper-
atures) is due to the large second-sphere contribution,
whereas the bell-shaped second part is essentially associated
with the inner-sphere contribution, which, in turn, seems to
be determined by a relatively long coordinated water ex-
change lifetime. Interestingly, the concomitant effect of a
high tM value for coordinated water molecules and the large
second-coordination-sphere contribution makes the overall
relaxivity of the Gd-AAZTAC17/defatted HSA adduct
almost independent of temperature.
Figure 8. 1/T1 NMRD profiles of a 1 mm solution of Gd-AAZTAC17
&
bound to defatted ( ) and fatted (&) HSA.
Gd-AAZTAC17 micelle/HSA supramolecular adduct: Bind-
ing to HSA has also been investigated for concentrations of
Gd-AAZTAC17 well above the critical micellar concentra-
tion (e.g., 1.3m m for the titration reported in Figure 10). In-
cal of systems endowed with long tr values. As both systems
also exhibit quite high relaxivities in the low frequency
range, the observed profiles suggest a large contribution to
the relaxivity arising from water protons in the second coor-
dination sphere of the chelate.
This contribution may be regarded as originating from
clusters of water molecules grafted in the proximity of the
chelate core just outside the hydrophobic pocket in which
the aliphatic chain is inserted. Analogous second-coordina-
tion-sphere contributions were previously reported for the
adducts formed upon non-covalent binding of a paramagnet-
ic complex at the surface of a protein[8] as well as in the
supramolecular “host–guest” adduct formed by poly-b-cy-
clodextrin and GdIII complexes.[44]
As confirmed by molecular docking studies on the acces-
sible solvent surface (see below), the difference in the relax-
ivities of fatted and defatted HSA adducts can be mainly as-
cribed to changes in the contribution of the second hydra-
tion sphere. Further support for the large contribution to
the relaxivity arising from the second hydration shell of the
complex bound to HSA comes from the variation of the lon-
gitudinal proton relaxation rates as a function of tempera-
ture. Values of rb1 (at 20 MHz) versus temperature are re-
ported in Figure 9. Experimental data may be interpolated
Figure 10. Plot of the variation of the observed longitudinal water proton
relaxation rate of a 1.3m m solution of Gd-AAZTAC17 as a function of
*
*
fatted ( ) and defatted ( ) HSA concentration (0.47 T, 258C).
terestingly, the behaviour shown by the micellar system was
quite different to the one observed when Gd-AAZTAC17
binds to HSA as a monomeric species.
Besides a slight increase in the relaxivity of the bound
form with respect to the free micellar system, very similar
values for the relaxivity of the bound forms were observed
with fatted and defatted HSA. This suggests that, above the
cmc, a supramolecular adduct is formed between the pre-
formed micelle and HSA. Such an association appears to be
driven by electrostatic interactions among the negative
charges of the micelles and the positively charged residues
exposed on the surface of the protein (Figure 11). In fact,
the increase in the ionic strength of the solution upon addi-
tion of NaCl invariantly caused a decrease in the observed
water proton relaxation rates.
Clearly, when the Gd-AAZTAC17 micelle is electrostati-
cally bound to HSA, any conformational change in the
structure of albumin brought about by the binding of fatty
acids does not interfere with the binding determinants that
are represented by positively charged amino acid residues.
Figure 9. Temperature dependence of the water proton relaxivity of the
Gd-AAZTAC17/defattedHSA adduct (0.47 T). The calculated contribu-
tions from water molecules in the inner, second and outer coordination
spheres to the observed relaxivity of the Gd-AAZTAC17/defatted HSA
supramolecular adduct are also reported.
Chem. Eur. J. 2007, 13, 5785 – 5797
ꢀ 2007 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chemeurj.org
5791

S. Aime et al.
and 2% Gd-AAZTAC17. Now the increase in relaxivity
brought about by this system (compared with the “all-Gd”
system) is only a few percentage units because the transloca-
tion to the protein of diamagnetic Y-AAZTAC17 units rules
out the contribution arising from the paramagnetic adduct
with HSA. Thus the observed relaxation enhancement ob-
served with Gd/Y mixed micelles comes only from the fac-
tors described above.
Molecular docking studies for the characterisation of the
binding site: In order to gain more insight into the differen-
ces in the binding mode of Gd-AAZTAC17 to fatted and
defatted HSA we carried out a molecular docking investiga-
tion. The competitive assays described above with palmitic
acid and warfarin clearly indicate that the binding region of
Gd-AAZTAC17 is represented by Sudlowꢁs site I.
Moreover it has been found that Gd-AAZTAC17 binds
with different affinities to fatted and defatted HSA as a con-
sequence of the conformational changes that occur upon the
binding of the fatty acids.[37] The molecular docking calcula-
tions allow us to evaluate in detail the effects these confor-
mational changes induce on the volume, polarity and solvent
accessibility[38] of the binding site I that may explain the re-
sults obtained in the relaxometric studies.
Figure 11. Schematic representation of the electrostatic interaction be-
tween the Gd-AAZTAC17 micelle and HSA.
As each micelle is formed by about 95 Gd-AAZTAC17
units, KA values of 1.5”105 and 1.7”105 mꢀ1 have been esti-
mated for the interaction of the micelle with defatted and
fatted HSA, respectively.
Further support for the supramolecular assembly formed
by the micelle and HSA has been gained by determining the
adduct size by the dynamic light scattering technique. The
mean diameter measured for the adduct (1:1 HSA/micelle)
was 11.5 nm, which nicely corresponds to the sum of the size
of albumin (experimental mean diameter ca. 6 nm) and the
size of the micelles (ca. 5.5 nm) (Figure 12).
The binding energies were estimated from calculations
using force field energy evaluation and thermal sampling by
molecular dynamic (MD) simulations. By comparing the ini-
tial structures and the average structures computed over the
last 50 ps trajectory of the MD simulation, rms coordinate
deviations were calculated for all heavy (non-hydrogen)
atoms within 16 Š of any ligand atoms. This yielded a rmsd
value of 1.1 Š for the adduct of Gd-AAZTAC17 and fatted
HSA and a value of 1.23Š for the corresponding one with
defatted HSA. The low rmsd values indicate that such simu-
lations are expected to reproduce interactions within the
binding region in a realistic fashion, that is, we can reasona-
bly assume that no unnatural distorting forces arise in the
modelling of these systems.
Energy data from the equilibration phase of the simula-
tions were not included in the interaction energy averages.
The data collection phase of the simulation was divided into
two portions and the average interaction energies were cal-
culated accordingly. The overall mean values were then de-
termined for each of the following terms: electrostatic, van
der Waals and the solvent-accessible surface area (SASA).
The average differences between the bound and free-state
simulations for each of the three terms are shown in
Table 3. DUele and DUvdw are the changes in the electrostatic
Figure 12. Comparative size distribution diagram, measured by dynamic
light scattering, of three independent solutions containing a) the Gd-
AAZTAC17 micelle ([Gd]=0.55 mm) (c), b) HSA ([HSA]=
0.0055 mm) (g) and c) the Gd-AAZTAC17 micelle/HSA adduct
([Gd]=0.55 mm, [HSA]=0.0055 mm) (a).
It is reasonable to surmise that the formation of the mi-
celle/HSA assembly is accompanied by the translocation of
Gd-AAZTAC17 monomers to the protein binding sites, as
described above. Therefore the observed relaxivity, though
mainly determined by gadolinium chelates organised in the
micellar form, also contains a contribution from the Gd-
AAZTAC17/HSA adduct. This may account for the slightly
higher relaxivity observed for the defatted HSA system
(Figure 10).
Table 3. Average differences in the ligand–surrounding interaction ener-
gies, as determined by MD simulations.
DUele
DUvdw
DSASA
[kcalmolꢀ1
]
[kcalmolꢀ1
]
[Š2]
Gd-AAZTAC17+fatted HSA
Gd-AAZTAC17+defatted HSA 158.312
121.234
ꢀ61.737
ꢀ55.425
ꢀ676.32
ꢀ626.32
Further support for this view was obtained by using the
above described micelles formed with 98% Y-AAZTAC17
5792
www.chemeurj.org
ꢀ 2007 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Chem. Eur. J. 2007, 13, 5785 – 5797

Binding of a Gd-AAZTA Complex to HSA
FULL PAPER
(coulombic) and van der Waals energies, respectively,
DSASA is the change in solvent-accessible surface area.
Molecular dynamics calculations have yielded DUvdw aver-
age energies for the binding of Gd-AAZTAC17 to fatted
HSA that are more negative than those with defatted HSA.
This is in accordance with the view that Gd-AAZTAC17
can interact more deeply inside the drug site I binding
cavity of fatted HSA.
If we consider the DSASA average for Gd-AAZTAC17,
the variation in solvent-accessible surface area is greater for
defatted HSA than for fatted HSA because, for defatted
HSA, the molecule can insert only a small portion of the
methylene tail inside the new
only half of its aliphatic tail into site I because Tyr 150 pre-
vents it from penetrating more deeply into the binding
cavity (Figure 13). The presence of Tyr 150 forces the Gd-
AAZTAC17 methylene tail to bend towards the hydropho-
bic residues of Leu 238, Val 241 and Leu 260.
In the case of fatted HSA the movement of Tyr 150 grants
access of the hydrophobic tail of Gd-AAZTAC17 to the
bottom of this channel. This is a hydrophobic subchamber,
delineated by Cys 245 and Cys 253(Figure 14). The hydro-
phobic tail of Gd-AAZTAC17 is involved in van der Waals
interactions with side-chain aliphatic carbon atoms of
channel arising from the rota-
tion of Tyr 150. This results in a
larger number of second-sphere
water molecules in close prox-
imity to the gadolinium chelate
bound to defatted HSA than is
the case with fatted HSA.
The absolute binding ener-
gies were estimated using
Equation (8) (see Experimental
Section), where a, b and g were
determined by fitting calculated
free energies of binding to ex-
perimental results. Values of
b=0.001, a=0.08 and g=
0.0026 yielded calculated bind-
ing energies of ꢀ6.57 and
ꢀ5.90 kcalmolꢀ1
for
Gd-
AAZTAC17/fatted and defat-
ted HSA adducts, respectively,
Figure 13. Graphical representation of Gd-AAZTAC17 bound to Sudlowꢁs site I on defatted HSA determined
compared with experimental by dynamic molecular modelling.
binding energies of ꢀ6.614 and
ꢀ5.947 kcalmolꢀ1.
We also investigated the
binding of Gd-AAZTAC17 to
fatted HSA at Sudlowꢁs site II,
obtaining a binding energy of
ꢀ5.02 kcalmolꢀ1. This value is
consistent with the experimen-
tal evidence obtained from
competition assays which re-
vealed that Gd-AAZTAC17
does not bind to drug site II.
These calculations showed
that upon binding of fatty
acids, Tyr 150 moves and inter-
acts with the carboxylate
moiety at fatty acid site II. This
structural change has a large
impact on drug site I, causing a
marked increase in the volume
of the binding pocket.[38]
Conversely, with defatted
HSA, Gd-AAZTAC17 can fit dynamic molecular modelling.
Figure 14. Graphical representation of Gd-AAZTAC17 bound to Sudlowꢁs site I on fatted HSA determined by
Chem. Eur. J. 2007, 13, 5785 – 5797
ꢀ 2007 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chemeurj.org
5793

S. Aime et al.
continued for 30 min after the addition. Then the reaction mixture was
cooled, filtered, washed with 10% aqueous Na2CO3 (3x100 mL) and
dried (Na2SO4). Removal of the solvent under reduced pressure gave 1-
nitrooctadecane (3) (3.9 g 71%). Colourless solid. M.p. 47–508C. 1H
NMR (CDCl3, 298 K): d=4.37 (t, J=6.9 Hz, 2H), 2.00 (q, J=7.1 Hz,
2H), 1.41–1.19 (m, 30H), 0.87 ppm (t, J=6.6 Hz, 3H); 13C NMR (CDCl3,
298 K): d=76.0 (CH2), 29.8 (8 CH2), 29.7 (CH2), 29.5 (CH2), 29.4 (CH2),
29.3(CH 2), 28.9 (CH2), 27.5 (CH2), 26.3(CH 2), 22.8 (CH2), 14.2 ppm
(CH3); ESI-MS: calcd for C18H37NO2: 299; found: m/z: 300 [M+H]+.
Lys 195, Lys 199 and Ala 291, whereas the last two methyl-
ene carbon atoms point towards Val 241.
Conclusion
The results reported in this work are relevant to the design
of gadolinium-based CAs for MRI for several reasons. First
of all it has been shown how the structural characteristics of
the binding site may affect the relaxation enhancement. In
fact it has been found that water molecules in the second co-
ordination sphere and exchangeable protons in the proximi-
ty of the interaction site may make an important contribu-
tion to the observed relaxivity. Moreover, the imbedding of
the metal chelate inside this kind of highly structured water
network causes a marked increase in the water exchange
rate at the paramagnetic centre. Thus it appears that two ef-
fects contribute in opposite ways to the observed relaxivity.
It is likely that the use of GdIII complexes characterised by a
very fast exchange of the coordinated water molecule may
provide a route to the attainment of very high relaxivities.
In fact, it seems reasonable to expect that the lengthening
caused by the embedding of such systems in the protein hy-
dration sphere would not yield the dramatically large tM
values that quench the attainable relaxivities. Alternatively,
the design of high relaxivity agents may result in the contri-
bution from the second coordination sphere being sacrificed
to maintain a fast exchange of the coordinated water. To
identify suitable binding sites in the latter systems, docking
simulations may be particularly advantageous.
It has also been shown that the presence of micelles may
dramatically alter the nature of the interaction of a lipophil-
ic agent with the target protein resulting in a much smaller
relaxation enhancement. This feature has to be taken into
serious consideration when a lipophilic agent is administered
in vivo under a highly concentrated bolus. Actually it ap-
pears likely that any self-assembled system bearing negative
charges on its surface would show this type of electrostatic
interaction with HSA. The formation of such a supramolec-
ular adduct may profoundly affect the biodistribution and
excretion pathways of the involved species.
1,4-Dibenzyl-6-heptadecyl-6-nitro-1,4-diazepane (4): N,N’-Dibenzylethy-
lenediamine diacetate (4.78 g, 13mmol) and 1-nitrooctadecane ( 3, 3.97 g,
13mmol) were dissolved in 1:1 toluene/methanol (100 mL) in a 500-mL
round-bottomed flask. Paraformaldehyde (1.39 g, 46.5 mmol) was added
portionwise to the resulting solution and the resulting suspension was re-
fluxed. The mixture became homogeneous (dissolution of paraformalde-
hyde) and after 3h at reflux, the mixture was cooled and evaporated in
vacuo. The residue was recrystallised from ethanol to obtain pure 4
(4.38 g, 61%) as a colourless solid. M.p. 70–728C (EtOH). 1H NMR
(CDCl3, 298 K): d=7.33–7.19 (m, 10H), 3.72 (d, J=13.0 Hz, 2H), 3.57
(d, J=13.0 Hz, 2H), 3.49 (d, J=14.2 Hz, 2H), 2.95 (d, J=14.2 Hz), 2.58
(m, 4H), 1.55 (m, 2H), 1.32–0.69 (m, 26H), 0.88 ppm (brt, 3H); 13C
NMR (CDCl3): d=139.3 (C), 129.2 (CH), 128.3 (CH), 127.3 (CH), 95.2
(C), 64.1 (CH2), 61.9 (CH2), 58.8 (CH2), 37.1 (CH2), 32.0 (CH2), 29.8 (7
CH2), 29.7 (CH2), 29.6 (CH2), 29.5 (2 CH2), 29.3(CH 2), 23.1 (CH2), 22.8
(CH2), 14.2 ppm (CH3); ESI-MS: calcd for C36H57N3O2: 563; found: m/z:
586 [M+Na]+, 564 [M+H]+.
6-Heptadecyl-1,4-diazepan-6-ylamine (5): Compound 4 (4.3g, 7.6 mmol)
was dissolved in methanol (100 mL) in a 250-mL flask. 10% Pd/C (1.0 g,
moistened with 0.5 mL water) was added followed by ammonium for-
mate (4.8 g, 76 mmol). The mixture was stirred and refluxed for 3h. The
catalyst was then removed by filtration and the filtrate evaporated in
vacuo. The residue was redissolved in dichloromethane and washed with
5% aq. NaOH. The organic phase was then dried (Na2SO4) to give 5
(2.6 g, 91%) as a colourless wax. M.p. 66–698C. 1H NMR (CDCl3,
298 K): d=2.87 (m, 4H), 2.72 (d, J=13.5 Hz, 2H), 2.61 (d, J=13.5 Hz,
2H), 1.77 (brs, 4H), 1.32–1.20 (m, 32H), 0.87 ppm (t, J=6.7 Hz, 3H);
13C NMR (CDCl3, 298 K): d=61.0 (CH2), 56.3(C), 52.2 (CH 2), 40.1
(CH2), 32.0 (CH2), 30.5 (CH2), 29.8 (10 CH2), 29.4 (CH2), 23.2 (CH2),
22.8 (CH2), 14.2 ppm (CH3); ESI-MS: calcd for C22H47N3: 353; found: m/
z: 376 [M+Na]+, 354 [M+H]+.
A
heptadecyl-1,4-diazepan-1-yl]acetic acid tert-butyl ester (6): The amine 5
(1.65 g 4.7 mmol) was dissolved in acetonitrile (15 mL) in a 50 mL round-
bottomed flask and K2CO3 (5.16 g, 37 mmol) was added. tert-Butyl bro-
moacetate (4.55 g, 23.4 mmol) was slowly added dropwise to the stirred
heterogeneous mixture while maintaining the temperature at <108C (ice
bath). After the addition the mixture was heated at 608C with stirring
until TLC showed complete conversion. The precipitate was filtered and
washed with dichloromethane; the filtrate and the washings were com-
bined and evaporated in vacuo to give the crude product. The semisolid
residue was purified by silica gel chromatography (petroleum ether/ethyl
acetate 9.3:0.7), giving pure 6 as a pale yellow oil (1.35 g, 36%). 1H
NMR (CDCl3, 298 K): d=3.63 (s, 4H), 3.22 (s, 4H), 2.99 (d, J=14.0 Hz,
2H), 2.75 (m, 4H), 2.64 (d, J=14.0 Hz, 2H), 1.44 (s, 18H), 1.43(s, 18H),
1.24 (m, 32H), 0.85 ppm (brt, 2H); 13C NMR (CDCl3, 298 K): d=173.0
(C), 170.9 (C), 80.7 (C), 80.2 (C), 65.4 (CH2), 63.2 (C), 62.6 (CH2), 59.4
(CH2), 52.0 (CH2), 37.7 (CH2), 32.0 (CH2), 30.6 (CH2), 29.9 (CH2), 29.8
(9 CH2), 29.4 (CH2), 28.3(CH 3), 28.2 (CH3), 22.8 (CH2), 22.1 (CH2),
14.2 ppm (CH3); ESI-MS: calcd for C46H87N3O8: 809; found: m/z: 83 2
[M+Na]+, 810 [M+H]+.
Finally it has been proved that the Gd-AAZTA complex
with its two coordinated water molecules is a structure of
considerable interest because of its ease of synthesis and the
possibility of functionalisation at the exocyclic carbon atom.
Remarkably, interaction with defatted HSA provides the
highest relaxivity reported so far for stable GdIII chelates.
Experimental Section
A
Synthesis of the AAZTAC17 ligand: Ligand AAZTAC17 was synthesised
following the reaction steps described in Scheme 1.
1-yl]acetic acid (1): The ester 6 (1.354 g 1.7 mmol) was dissolved in tri-
fluoroacetic acid (20 mL) in a 50-mL round-bottomed flask and stirred at
room temperature overnight. The solution was then evaporated in vacuo
and the residue was redissolved in methanol (2 mL). The product precipi-
tated from solution on addition of excess diethyl ether, was isolated by
centrifugation, washed thoroughly with diethyl ether and dried in vacuo,
giving pure 1 (795 mg, 81%) as an amorphous white solid. M.p. 185–
1-Nitrooctadecane (3): 3-Chloroperoxybenzoic acid (12.84 g, 74.4 mol,
85% pure) was dissolved in 1,2-dichloroethane (200 mL) in a two-necked
flask equipped with a condenser and a pressure-equalising dropping
funnel. Octadecylamine (2, 5.0 g, 18.6 mmol) in 1,2-dichloroethane
(100 mL) was added dropwise to the refluxing solution. Refluxing was
5794
www.chemeurj.org
ꢀ 2007 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Chem. Eur. J. 2007, 13, 5785 – 5797

Binding of a Gd-AAZTA Complex to HSA
FULL PAPER
1878C (decomp). 1H NMR (CD3OD, 298 K): d=3.85 (d, J=17.8 Hz,
2H), 3.80 (d, J=17.8 Hz, 2H), 3.71 (s, 4H), 3.35 (d, J=14.3Hz, 2H),
3.34 (s, 4H), 3.24 (d, J=14.3 Hz, 2H), 1.39 (m, 2H), 1.28 (m, 30H),
0.89 ppm (t, J=6.7 Hz, 3H); 13C NMR (CD3OD, 298 K): d=175.8 (C),
170.0 (C), 62.5 (C), 60.5 (CH2), 57.8 (CH2), 53.1 (CH2), 50.3(CH 2), 35.5
Minimisation was achieved by a multistep procedure, first fixing all
heavy atoms and relaxing the positions of the added hydrogen atoms,
then through geometry optimisation of the hydrogen atoms and side-
chains while keeping the backbone atoms fixed, and finally through full
geometry optimisation without constraints until convergence was less
ꢀ1
(CH2), 31.8 (CH2), 30.0 (CH2), 29.5 (9 CH2), 29.3(CH ), 29.2 (CH2), 23.1
than 0.01 kcalmolꢀ1
methods.
Š
with steepest-descent and conjugate gradient
2
(CH2), 22.5 (CH2), 13.2 ppm (CH3); ESI-MS: calcd for C30H55N3O8: 585;
found: m/z: 608 [M+Na]+, 586 [M+H]+.
The ligand structure of Gd-AAZTA was constructed starting from the
crystal structure of the AAZTA-like cycle obtained from the CSD
(RULLEP code) with the Moe-Builder module and optimised by ab
initio methods. Full geometry optimisation was performed at the RHF
level with Gaussian 98[46] using an effective core potential (ECP) that in-
cludes 4f electrons in the core (1s–4d,4f7)[47] and an optimised valence
basis set for the metal and 3-21G basis sets for ligand atoms.[48] The meth-
ylene tail was added with the Moe-Builder module and minimised keep-
ing the AAZTA cycle fixed.
Complexation of AAZTAC17 with GdIII: The GdIII complex was readily
synthesised by adding stoichiometric amounts of the ligand and GdCl3 at
room temperature. The excess free gadolinium was removed by filtering
the basic solution with 0.2 mm syringe filters. Orange Xylenol UV spec-
trophotometry was used to check for the absence of free GdIII ions.[45]
Water proton relaxivity measurements: The longitudinal water proton re-
laxation rates were measured by using a Stelar Spinmaster (Mede, Pavia,
Italy) spectrometer operating at 0.47 T by the standard inversion-recov-
ery technique (16 experiments, 2 scans). A typical 908 pulse width was
3.5 ms and the reproducibility of the T1 data was ꢁ0.5%. The tempera-
ture was controlled with a Stelar VTC-91 air-flow heater equipped with a
copper/constantan thermocouple (uncertainty ꢁ0.18C). The proton 1/T1
NMRD profiles were measured over a continuum of magnetic field
strength from 0.00024 to 0.47 T (corresponding to a 0.01–20 MHz proton
Larmor frequency) on a Stelar field-cycling relaxometer. The relaxome-
ter operates under complete computer control with an absolute uncer-
tainty in 1/T1 of ꢁ1%. Data points from 0.47 (20 MHz) to 1.7 T
(70 MHz) were collected on a Stelar Spinmaster spectrometer operating
at variable fields.
17O measurements: Variable-temperature 17O NMR measurements were
recorded with a JEOL EX-90 spectrometer equipped with a 5 mm probe
using a D2O external lock. Experimental settings were as follows: spec-
tral width 10000 Hz, 908 pulse width (7 ms), acquisition time 10 ms, 1000
scans and no sample spinning. Solutions containing 2.6% of the 17O iso-
tope (Yeda, Israel) were used. The observed transverse relaxation rates
(R02obs) were calculated from the signal width at half-height (Dn1/2):
R02obs =pDn1/2
Partial charges were calculated by the Mulliken method implemented in
Gaussian 98 at the RHF level of theory using the 6-31G* basis set for
ligand atoms and the ECP for the gadolinium atom with a pseudopoten-
tial according to Dolg and Stoll.[48] The calculated charges were averaged
over all atoms of the same type, with the metal ion maintaining a +3
atomic charge[49] and the total charge was then corrected by distributing
the excess charge onto donor atoms.
For all the calculations a modification of the Amber99 force field[50] was
used with in-house parametrisation to treat gadolinium complexes within
the framework of the ionic method.[51] The s and e van der Waals param-
eters for Gd3+ were chosen to reproduce the experimental (crystallo-
graphic) coordination structures of the Gd-DOTA and Gd-DTPA com-
plexes (CSD code: JOPJIH and HEQBUA, respectively). This was ach-
ieved by energy minimisation and by systematically changing the value of
s and e between each minimisation in order to reproduce the crystallo-
graphic geometry (lower RMSD values).[52]
The method was evaluated using known structures, obtaining the best fit
to crystal structures with s=1.8 Š and e=0.1, giving rms deviations of
0.32 Š for Gd-DOTA and 0.31 Š for Gd-DTPA.
Light scattering measurements: Dynamic light scattering measurements,
made to determine the size of the micellar system, were performed on a
Malvern Zetasizer Nano ZS apparatus. The size range over which the in-
strument can operate is 0.6 nm–6 m. A laser was used as the light source
to illuminate the sample particles within a cell. A detector, positioned at
1738, was used to measure the intensity of the light scattered by the parti-
cles of the sample. An attenuator was used to reduce the intensity of the
laser and hence reduce the intensity of the scattering which has to be
within a specific range for the detector to successfully measure it. The po-
sition of the attenuator was determined automatically by the Zetasizer
during the measurement process. The scattering intensity detected was
passed to a digital signal processing board called a correlator. The corre-
lator compares the scattering intensity at successive time intervals to
derive the rate at which the intensity varies. This correlator information
was then passed to a computer where the specialist Zetasizer software
analyses the data and derives size information.
The starting positions for the docking procedure were obtained by modi-
fying the ligand positions, orientations and conformations to optimise
binding geometry while filling the available space in the HSA drug site I.
The docking procedure was performed using the Moe-Dock module with
Tabu Search with 10 runs, 1000 steps per run. The ligand was kept partial-
ly flexible.
The best solution from the docking calculations was submitted to molecu-
lar dynamics simulations of the bound and unbound ligand, the solvating
ligand and protein–ligand complexes with 20 Š spheres of TIP3P water
molecules. For the simulations of the ligand bound to the protein, resi-
dues within 20 Š of the binding site were included; the outer residues
were kept frozen. The simulation of the free state was carried out in a
sphere of equal size filled with water molecules and with the ligand at
the centre.
In both simulations the water molecules were restrained to prevent va-
porisation using a half-harmonic restraint 20 Š from the ligand atoms
Molecular weight measurements of the Gd-AAZTAC17 micellar system
were managed using the molecular weight function in the DTS software
of the Zetasizer Nano system, which compiles the static intensity meas-
urements, generates a standard Debye plot and then calculates the mo-
ꢀ2
with a force constant of 20 kcalmolꢀ1
Š
.
To have the same net charge in the bound and free states to avoid Born-
type corrections,[53] excess charge on the protein must be removed. To
fulfil this requirement, charged residues far from the ligand and near the
boundary of the simulation sphere were neutralised.[54]
lecular weight and second virial coefficient. Molecular weight measure-
[27,28]
ments were performed in water using a dn/dc value of 0.12 mLgꢀ1
.
Different sample concentrations were prepared by diluting a high con-
centration stock solution and filtering through 100 nm Anotop filters
prior to measurement. Both the toluene reference and solvent were
double-filtered through 20 nm Anotop filters prior to measurement.
Molecular dynamics simulations were performed with the NVT ensemble
keeping the temperature constant at 308C, as in the binding experiment.
After minimisation of the water sphere until
a rms gradient of
0.01 kcalmolꢀ1 Š was obtained, the temperature was increased to 300 K,
followed by 20 ps of equilibration of water molecules keeping the ligand
fixed, then 20 ps of equilibration of the entire system. Data were collect-
ed during the next 100 ps of the simulations.
Computational methods: All computer modelling procedures were car-
ried out using the Moe program (Chemical Computing Group, Inc.,
MOE 2004.03(www-chemcomp.com). The high-resolution three-dimen-
sional coordinates of human serum albumin (HSA), fat-free and com-
plexed with stearic acid, were obtained from the Protein Data Bank
(PDB code 2bcx and 1E7I). The structures were prepared for docking
calculations by adding hydrogen atoms and completing missing atoms.
The free energies of binding were estimated using a recently reported
method proposed and successfully tested on different endothiapepsin in-
hibitor complexes by Aqvist et al.[53] This method is based on standard
thermodynamic cycles and on linear approximation of polar and non-
Chem. Eur. J. 2007, 13, 5785 – 5797
ꢀ 2007 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chemeurj.org
5795

S. Aime et al.
polar free energy contributions from the corresponding electrostatic and
van der Waals energy averages. The free energy of binding DGb of a
ligand is approximated by Equation (8), where the DU energies are, re-
spectively, the electrostatic and van der Waals energy differences be-
tween the ligand in the solvated protein (bound state) and the ligand in
the sphere of the water molecules (free state).
[15] S. Aime, M. Botta, S. Geninatti Crich, G. B. Giovenzana, G. Jommi,
R. Pagliarin, M. Sisti, Inorg. Chem. 1997, 36, 2992–3000.
[16] S. Aime, M. Botta, S. Geninatti Crich, G. B. Giovenzana, G. Jommi,
R. Pagliarin, M. Sisti, J. Chem. Soc., Chem. Commun. 1995, 1885–
1886.
[17] S. Aime, E. Gianolio, D. Corpillo, C. Cavallotti, G. Palmisano, M.
Sisti, G. B. Giovenzana, Helv. Chim. Acta 2003, 86, 615–632.
[18] J. Xu, S. J. Franklin, Jr., D. W. Whisenhunt, K. N. Raymond, J. Am.
Chem. Soc. 1995, 117, 7245–7246.
DGb ¼ b < DUele > þa < DUvdw > þg < DSASA >
ð8Þ
[19] S. Hajela, M. Botta, S. Giraudo, J. Xu, K. N. Raymond, S. Aime, J.
Am. Chem. Soc. 2000, 122, 11228–11229.
[20] D. M. J. Doble, M. Botta, J. Wang, S. Aime, A. Barge, K. N. Ray-
mond, J. Am. Chem. Soc. 2001, 123, 10758–10759.
[21] S. Aime, L. Calabi, C. Cavallotti, E. Gianolio, G. B. Giovenzana, P.
Losi, A. Maiocchi, G. Palmisano, M. Sisti, Inorg. Chem. 2004, 43 ,
7588–7590.
[22] S. H. Koenig, R. D. Brown, Prog. Nucl. Magn. Reson. Spectrosc.
1990, 26, 487–567.
[23] D. H. Powell, O. M. Ni Dhubhghaill, D. Pubanz, L. Helm, Y. S. Leb-
edev, V. Schlaepfer, A. E. Merbach, J. Am. Chem. Soc. 1996, 118,
9333–9346.
Furthermore, the binding free energy is not determined by a single con-
formation from the docking procedure, but indeed by a thermal average.
Therefore this approach tries to capture conformational fluctuations that
are important for the binding energy.
Interaction energies for the ligand in the binding pocket of the solvated
protein and in the aqueous environment were extracted using a 16 Š cut-
off value.
Solvent accessible surface area calculations were carried out after the
simulations on coordinates recorded every 0.5 ps of the sampling phase.
Average energies and SASA differences were fitted to the experimental
DG binding values to obtain linear response parameters a, b and g ac-
cording to Equation (8). This procedure was performed with a genetic al-
gorithm. Experimental DG values were calculated from KA values report-
ed in Table 2 (DG=ꢀRTlnKA).
[24] S. Aime, M. Botta, M. Fasano, E. Terreno, Acc. Chem. Res. 1999, 32,
941–949.
[25] G. M. Nicolle, E. Toth, K. P. Eisenwiener, H. R. Macke, A. E. Mer-
bach, J. Biol. Inorg. Chem. 2002, 7, 757–769.
[26] S. Laus, A. Sour, R. Ruloff, E. Toth, A. E. Merbach, Chem. Eur. J.
2005, 11, 3064–3076.
[27] G. M. Nicolle, L. Helm, A. E. Merbach, Magn. Reson. Chem. 2003,
41, 794–799.
[28] H. G. Thomas, A. Lomakin, D. Blankschtein, G. B. Benedek, Lang-
muir 1997, 13, 209–218.
Acknowledgements
[29] V. Schadler, C. Nardin, U. Wiesner, E. Mendes, J. Phys. Chem. B
2000, 104, 5049–5052.
[30] P. Debye, J. Appl. Phys. 1944, 15, 338–343.
Support from Bracco Imaging Spa, PRIN 2005, Noe-EMIL and DiMI is
gratefully acknowledged. This work was carried out within the frame-
work of EU-Cost D18 Action.
[31] P. J. Wyatt, Anal. Chim. Acta 1993, 272, 1–40.
[32] K. Mattison, M. Kaszuba, American Biotechnology Laboratory, June
2003.
[33] U. Kragh-Hansen, Dan. Med. Bull. 1990, 37, 57–84.
[34] D. C. Carter, J. X. Ho, Adv. Protein Chem. 1994, 47, 152–203.
[35] T. J. Peters, All about Albumin: biochemistry, genetics and medical
applications, Academic Press, San Diego, 1996.
[36] S. Curry, P. Brick, N. P. Franks, Biochim. Biophys. Acta 1999, 1441,
131–140.
[37] S. Curry, H. Mandelkov, P. Brick, N. Franks, Nat. Struct. Biol. 1998,
5, 827–835.
[38] J. Ghuman, P. A. Zunszain, I. Petitpas, A. A. Bhattacharya, M. Ota-
giri, S. Curry, J. Mol. Biol. 2005, 353, 38–52.
[39] J. B. Livramento, C. Weidensteiner, M. I. M. Prata, P. R. Allegrini,
C. F. G. C. Geraldes, L. Helm, R. Kneuer, A. E. Merbach, A. C.
Santos, P. Schmidt, E. Toth, Contrast Media Mol. Imaging 2006, 1,
30–39.
[40] X. M. He, D. C. Carter, Nature 1992, 358, 209–215.
[41] R. Reed, J. Biol. Chem. 1977, 252, 7483–7487.
[42] H. Vorum, B. Honorꢄ, J. Pharm. Pharmacol. 1996, 48, 870–875.
[43] S. Sugio, A. Kashima, S. Mochizuki, M. Noda, K. Kobayashi, Protein
Eng. 1999, 12, 439–446.
[1] R. B. Lauffer, Chem. Rev. 1987, 87, 901–927.
[2] A. E. Merbach, E. Toth, The Chemistry of Contrast Agents in Medi-
cal Magnetic Resonance Imaging, Wiley, Chichester, 2001.
[3] S. Aime, M. Botta, M. Fasano, E. Terreno, Chem. Soc. Rev. 1998, 27,
19–29.
[4] P. Caravan, J. J. Ellison, T. J. McMurry, R. B. Lauffer, Chem. Rev.
1999, 99, 2293–2352.
[5] H. J. Weinmann, A. Mꢂhler, B. Radꢂchel in Biomedical Magnetic
Resonance Imaging and Spectroscopy (Ed.: I. R. Young), Wiley, Chi-
chester, 2000, p. 705.
[6] L. Banci, I. Bertini, C. Luchinat, Nuclear and Electronic Relaxation,
VCH, Weinheim, 1991.
[7] S. Aime, A. Barge, E. Gianolio, S. Geninatti Crich, W. Dastrꢃ, F.
Uggeri in Metallotherapeutic drugs
& metal based Diagnostic
Agents; The use of Metals in Medicine, (Eds.: M. Gielen, E. R. T.
Tiekink), Wiley, 2005, p.541.
[8] S. Aime, M. Botta, M. Fasano, E. Terreno, Protein-bound metal che-
lates in The chemistry of contrast agents in medical Magnetic Reso-
nance Imaging. (Eds.: A. E. Merbach, E. Toth), Wiley, Chichester,
2001, p. 193.
[44] S. Aime, M. Botta, F. Fedeli, E. Gianolio, E. Terreno, P. L. Anelli,
Chem. Eur. J. 2001, 7, 5261–5269.
[9] M. H. Ou, C. H. Tu, S. C. Tsai, W. T. Lee, G. C. Lin, Y. M. Wang,
Inorg. Chem. 2006, 45, 244–254.
[45] A. Barge, G. Cravotto, E. Gianolio, F. Fedeli, Contrast Media Mol.
Imaging 2006, 1, 184–188.
[10] S. Aime, M. Chiaussa, G. Digilio, E. Gianolio, E. Terreno, J. Biol.
Inorg. Chem. 1999, 4, 766–774.
[46] Gaussian 98, Revision A.11, M. J. Frisch, G. W. Trucks, H. B. Schle-
gel, G. E. Scuseria, M. A. Robb, J. R. Cheeseman, V. G. Zakrzewski,
J. A. Montgomery, Jr., R. E. Stratmann, J. C. Burant, S. Dapprich,
J. M. Millam, A. D. Daniels, K. N. Kudin, M. C. Strain, O. Farkas, J.
Tomasi, V. Barone, M. Cossi, R. Cammi, B. Mennucci, C. Pomelli,
C. Adamo, S. Clifford, J. Ochterski, G. A. Petersson, P. Y. Ayala, Q.
Cui, K. Morokuma, P. Salvador, J. J. Dannenberg, D. K. Malick,
A. D. Rabuck, K. Raghavachari, J. B. Foresman, J. Cioslowski, J. V.
Ortiz, A. G. Baboul, B. B. Stefanov, G. Liu, A. Liashenko, P. Piskorz,
[11] S. Aime, E. Gianolio, D. Longo, R. Pagliarin, C. Lovazzano, M.
Sisti, ChemBioChem 2005, 6 , 818.
[12] F. A. Dunand, A. Borel, A. E. Merbach, J. Am. Chem. Soc. 2002,
124, 710–716.
[13] S. Aime, E. Gianolio, E. Terreno, G. B. Giovenzana, R. Pagliarin,
M. Sisti, G. Palmisano, M. Botta, M. Lowe, D. Parker, J. Biol. Inorg.
Chem. 2000, 5, 488–497.
[14] S. Aime, M. Botta, S. Geninatti Crich, G. B. Giovenzana, R. Pagliar-
in, M. Sisti, E. Terreno, Magn. Reson. Chem. 1998, 36, 200–208.
5796
www.chemeurj.org
ꢀ 2007 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Chem. Eur. J. 2007, 13, 5785 – 5797

Binding of a Gd-AAZTA Complex to HSA
FULL PAPER
I. Komaromi, R. Gomperts, R. L. Martin, D. J. Fox, T. Keith, M. A.
Al-Laham, C. Y. Peng, A. Nanayakkara, M. Challacombe, P. M. W.
Gill, B. Johnson, W. Chen, M. W. Wong, J. L. Andres, C. Gonzalez,
M. Head-Gordon, E. S. Replogle, J. A. Pople, Gaussian, Inc., Pitts-
burgh, PA, 2001.
[50] J. W. Ponder, D. A. Case, Adv. Protein Chem. 2003, 66, 27–35.
[51] B. P. Hay, Coord. Chem. Rev. 1993, 126, 177–236.
[52] R. Fossheim, S. G. Dahl, Acta Chem. Scand. 1990, 44, 698–706.
[53] J. Aqvist, C. Medina, J. E. Samuelsson, Protein Eng. 1994, 7, 385–
391.
[47] U. Cosentino, G. Moro, D. Pitea, A. Villa, P. C. Fantucci, A. Maioc-
chi, F. Uggeri, J. Phys. Chem. A 1998, 102, 4606–4614.
[54] J. Aqvist, J. Comput. Chem. 1996, 17, 1587–1597.
[48] M. Dolg, H. Stoll, Theor. Chim. Acta 1989, 75, 173–179.
[49] F. Yerly, K. I. Hardcastle, L. Helm, S. Aime, M. Botta, A. E. Mer-
bach, Chem. Eur. J. 2002, 8, 1031–1039.
Received: September 5, 2006
Published online: April 3, 2007
Chem. Eur. J. 2007, 13, 5785 – 5797
ꢀ 2007 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chemeurj.org
5797