Chemical Characterization of α-Oxohydrazone Ligation on Colloids: Toward Grafting Molecular Addresses onto Biological Vectors

Article abstract of DOI:10.1021/ja0370746

New mild and specific chemical strategies have been developed recently for the selective coupling of biological macromolecules. Among them, the hydrazone ligation strategy offers high chemoselectivity and versatility. We intended to use hydrazone ligation to target the controlled release of therapeutic agents by biological vectors (multilamellar vesicles called onion vectors). An accurate measure of ligation bond stability was needed to ensure that the ligation bond would stand long exposures to physiological conditions. In this study, we have completed a kinetic and thermodynamic characterization of hydrazone formation on a model reaction. The mechanism of the reaction in solution as well as in different self-organized systems (micelles, liposomes and multilamellar vesicles) was investigated. In solution, submicromolar stability was achieved as well as half-lives of several weeks. The kinetics and stability were both enhanced in colloidal media thanks to autoassociation effects. The results were expanded to the realistic case of RGD-peptide coupling to onion vectors. The RGD grafted onion vectors were then tested for their ability to bind endothelial cells in vitro.

Full text of DOI:10.1021/ja0370746

Published on Web 11/27/2003
Chemical Characterization of r-Oxohydrazone Ligation on
Colloids: toward Grafting Molecular Addresses onto
Biological Vectors
P. Chenevier,*,†,§ L. Bourel-Bonnet,‡ and D. Roux†
Contribution from the Centre de Recherche Paul Pascal, CNRS UPR 8641, aV. Pr Schweitzer,
33600 Pessac, France, and Institut de Biologie de Lille, UMR 8525, 1, rue Calmette, 59021
Lille Cedex, France
Received July 4, 2003; E-mail: pac38@cornell.edu
Abstract: New mild and specific chemical strategies have been developed recently for the selective coupling
of biological macromolecules. Among them, the hydrazone ligation strategy offers high chemoselectivity
and versatility. We intended to use hydrazone ligation to target the controlled release of therapeutic agents
by biological vectors (multilamellar vesicles called onion vectors). An accurate measure of ligation bond
stability was needed to ensure that the ligation bond would stand long exposures to physiological conditions.
In this study, we have completed a kinetic and thermodynamic characterization of hydrazone formation on
a model reaction. The mechanism of the reaction in solution as well as in different self-organized systems
(micelles, liposomes and multilamellar vesicles) was investigated. In solution, submicromolar stability was
achieved as well as half-lives of several weeks. The kinetics and stability were both enhanced in colloidal
media thanks to autoassociation effects. The results were expanded to the realistic case of RGD-peptide
coupling to onion vectors. The RGD grafted onion vectors were then tested for their ability to bind endothelial
cells in vitro.
1. Introduction
on natural macromolecules because they contain neither hydra-
zine nor reactive aldehyde. Finally, the hydrazine and aldehyde
moieties can be specifically designed onto synthetic as well as
natural macromolecules.6 An important question that remains
concerns the kinetic and thermodynamic stability of the R-oxo-
hydrazone bond. Indeed, hydrazones are known to hydrolyze
quite easily in water, whereas ligation products such as
immunotoxins or grafted vectors should be able to face
extremely dilute aqueous conditions during hours if they were
to be used as therapeutic agents. Fast hydrolysis is a major
hurdle for using ligation in biological applications. While this
question has been addressed for other systems such as disulfide
bonds,2 both the thermodynamic stability and kinetics of the
R-oxohydrazone bond have to be accurately understood before
it can be trusted as a good ligation strategy.
Onion vectors are multilamellar vesicles which differ from
liposomes in the structure and method of preparation. Obtained
by shearing of lipid lamellar phases7 (a simple process easily
scalable to industrial volumes), onion vectors can encapsulate
large amounts of proteins, DNA, or organic molecules in good
yields.8 The shearing process turns the lamellar phase into a
compact stack of multilamellar vesicles. Depending on the
choice of lipids and surfactants, this phase can be dispersed in
With the recent progress in recombinant protein synthesis and
chemical peptide and protein synthesis, complex macro-
molecules have been increasingly available for chemical modi-
fication and therapeutic uses. The manipulation of such large
and delicate molecules is a new chemical challenge though and
requires particular tools. Indeed, any chemical modification must
be achieved in mild aqueous conditions to avoid denaturation
of the macromolecule. Furthermore, for large polymers pos-
sessing multiple equivalent reactive moieties, the regioselectivity
of chemical reaction has to be precisely controlled. A new series
of chemical reactions addressing these requirements have been
recently developed and designated as ligation reactions. Liga-
tions can be used for large protein chemical synthesis,1 for
macromolecule coupling as in the case of immunotoxins,2 or
for grafting molecular addresses onto biological vectors.3
The R-oxohydrazone bond formation is well suited for this
purpose. First, the reaction is done in mild aqueous conditions,
and no undesirable subproduct is produced. Second, R-oxo-
hydrazone compounds are obtained in good yields and with high
chemoselectivity.4,5 Third, the reaction is very regioselective
† Centre de Recherche Paul Pascal.
‡ Institut de Biologie de Lille.
(4) Melnyk, O.; Bossus, M.; David, D.; Rommens, C.; Gras-Masse, H. J. Pept.
Res. 1998, 52, 180-184.
§ Present address: LASSP, Cornell University, Ithaca, NY 14853, USA.
(1) Nilsson, B. L.; Hondal, R. J.; Soellner, M. B.; Raines, R. T. J. Am. Chem.
Soc. 2003, 125, 5268-5269.
(5) Bonnet, D.; Bourel, L.; Gras-Masse, H.; Melnyk, O. Tetrahedron Lett. 2000,
41, 10003-10007.
(2) Arpicco, S.; Dosio, F.; Brusa, P.; Crosasso, P.; Cattel, L. Bioconjugate
Chem. 1997, 8, 327-337.
(6) Melnyk, O.; Fehrentz, J.-A.; Martinez, J.; Gras-Masse, H. Biopolymers 2000,
55, 165-186.
(3) Kirpotin, D.; Park, J. W.; Hong, K.; Zalipsky, S.; Li, W.-L.; Carter, P.;
Benz, C. C.; Papahadjopoulos, D. Biochemistry 1997, 36, 66-75.
(7) Diat, O.; Roux, D.; Nallet, F. J. Phys. II France 1993, 3, 1427-1452.
(8) Pott, T.; Roux, D. FEBS Lett. 2002, 511, 150-154.
9
10.1021/ja0370746 CCC: $25.00 © 2003 American Chemical Society
J. AM. CHEM. SOC. 2003, 125, 16261-16270
16261

A R T I C L E S
Chenevier et al.
spectroscopy (TOF MS) analysis gave for RGDH and RGEH 1139.4
(MALDI) and 1156 (ESI), for calculated mass of protonated species
1139.3 and 1154.4, respectively.
Lipophilic R-oxoaldehyde (LA) was prepared as described previ-
ously.19 Water soluble R-oxoaldehyde (WA) was kindly provided by
Nathalie Ollivier: peptide SKYVL-NH2 was prepared by automatic
solid-phase peptide synthesis and oxidized by periodate oxidation.20
Lipid-ligands LRGD and LRGE were prepared and used as described
previously.14,18
Figure 1. Grafting of ligands (hand motifs) onto onion vectors using the
R-oxohydrazone ligation; transmission electron microscopy image from ref
16.
The composition of common buffers were Tris 10 mM/EDTA 1 mM
at pH 7.4 and acetic acid 100 mM at pH 4.6. Other buffers were
hydrochloric (HCl 200 mM + KCl 200 mM, pH 1 to 2), citrate-
phosphate (citric acid 50 mM + Na2HPO4 100 mM, pH 2.6 to 7), acetic
(acetic acid 100 mM + NaOH, pH 3.6 to 5), carbonate (K2CO3 50
mM + HCl, pH 8 to 9.6), and phosphate (Na2HPO4 50 mM + Na3PO4
50 mM, pH 10 to 12) buffers.
2.2. Micelles, Liposomes, and Onion Vectors. For micelle prepara-
tions, LA dissolved in ethanol at 60 °C was mixed with surfactant C12E7
and finally dispersed in excess buffer. Final concentrations were 25 to
100 µM in LA, 0.25 to 1 wt % (5 to 21 mM) in C12E7 and less than 2%
ethanol.
For liposomes and onion vectors, lipids were mixed in ethanol at
45 °C, and the ethanol was evaporated under vacuum, prior to the
addition of pure water to achieve the final weight compositions: PC/
Sim/water 45:20:35 for control and LA/PC/Sim/water 6.5:38.5:20:35
for LA onion vectors. Samples were incubated over half a day, before
the lamellar phase was sheared between the cone tube walls and a
matching cone pestle. The onion vector suspension was produced by
adding a 10 time excess of pure water. For liposomes, the previous
stock suspension was diluted 10 times in buffer and sonicated at high
power for 10 min (10 W in 1 mL sample). Metal particles from the
sonication probe were removed by a short centrifugation.
2.3. Determination of Reaction Rate Constants by Fluorometry.
Rates of reaction with MBTH were measured by recording the
fluorescence of the sample, after adding MBTH to a solution or colloidal
suspension containing the aldehyde reagent A (WA or LA) to produce
product P:
excess water. A suspension of multilamellar vesicles with
concentric membranes up to the core is obtained.9 They have
proved efficient as vectors for the transport of active compounds
in vivo as well as in vitro,10,11 especially for vaccine delivery.12,13
We have recently described their first applications as neutral
targeted vectors. First, we used RGD ligands attached to a lipid
moiety to ensure their association to the onion vectors.14 The
strategy was successful; however, the synthesis of the lipid-
ligands proved to be delicate. Therefore, we turned to a ligation
strategy, using reactive synthetic lipids in the onion vectors to
graft mannose mimetic ligands in situ15 (Figure 1). In the present
study, we intend to fully understand the ligation process, so
that the grafting onto onion vectors can be easily adapted to
any type of ligands.
Here, we report a complete characterization of the reaction
of R-oxohydrazone bond formation, in aqueous solution as well
as in self-organized media. As R-oxohydrazone bonds were
difficult to detect in complex media, we first studied the
mechanism, kinetics, and thermodynamics of the reaction using
closely related model reagents. We then used optimal conditions
for the grafting of targeting ligands, RGD peptides, onto onion
vectors, and tested the biological efficiency of the grafted vectors
on target cells in vitro.
2. Experimental Section
2.1. Reagents and Chemicals. The buffer compounds, 3-methyl-
2-benzothiazolinone hydrazone (MBTH) hydrochloride, dioleoyltri-
methylaminomethane (DOTAP), and polyoxyethylene 7 lauryl ether
(C12E7) were provided by Sigma, soybean phosphatidylcholine S100
(PC) was provided by Lipoid (France), and polyoxyethylene 8
stearoylester or Simulsol2599 (Sim) was provided by SEPPIC (France).
Hydrazinoacetyl peptides RGDH and RGEH were synthesized by
solid-phase peptide synthesis, with Fmoc/tert-butyl standard procedures
and hydrazine coupling methods developed by Bonnet et al.17 Briefly,
the YGRGDSP and YGRGESP sequences were synthesized automati-
cally.18 A glutaric moiety, 4,7,10-trioxa-1,13-tridecanediamine, and
N,N′,N′-tri(tert-butyloxycarbonyl)hydrazinoacetic acid were coupled
manually. The products were then cleaved and purified by RP-HPLC
with 61% (RGDH) and 56% (RGEH) overall yield. The mass
A + MBTH
9
k8 P + H2O
(1)
where k is the global reaction rate constant. The rate constant of the
reverse reaction was supposed to be negligible compared to k, a
hypothesis justified by the high stability of P. Fluorescence intensity
was measured in a thermostated 5 mm square quartz cuvette, using a
Fluoromax (Spex, USA) fluorometer. The fluorescence intensity
(excitation 366 nm, emission 495 nm) was linear in the concentration
of P up to 60 µM and 100 µM for the products of reaction of MBTH
with WA and LA, respectively (data no shown). Fluorescence intensity
time series were then fitted to the second-order reaction rate equation:
1 - e-t/τ
IF ) IF0 + IFmax
(2)
1 - (a/n)e-t/τ
(9) Gulik-Krzywicki, T.; Dedieu, J. C.; Roux, D.; Degert, C.; Laversanne, R.
Langmuir 1996, 12, 4668-4671.
where τ ) k-1(n - a)-1, IF, IF0, and IFmax are the recorded
instantaneous, initial, and final fluorescence intensities, respectively,
and a and n, the initial global concentrations of A and MBTH (τ, IF0,
and IFmax were the adjustable parameters). k was determined from
several values of τ measured with different values of n and a.
Concentrations ranged typically from 200 µM to 2 mM for MBTH and
between 5 and 100 µM for R-oxoadehydes.
(10) Freund, O.; Ame´de´e, J.; Roux, D.; Laversanne, R. Life Sci. 2000, 67, 411-
419.
(11) Mignet, N.; Brun, A.; Degert, C.; Delord, B.; He´le`ne, C.; Laversanne, R.;
Franc¸ois, J.-C. Nucleic Acids Res. 2000, 28, 3134-3142.
(12) Gaubert, S. World Patent WO99/16468, 1998.
(13) Gaubert, S. World Patent WO01/19335A2, 2001.
(14) Chenevier, P.; Delord, B.; Ame´de´e, J.; Bareille, R.; Ichas, F.; Roux, D.
Biochim. Biophys. Acta 2002, 1593, 17-27.
(15) Chenevier, P.; Grandjean, C.; Loing, E.; Malingue, F.; Angyalosi, G.; Gras-
Masse, H.; Roux, D.; Melnyk, O.; Bourel-Bonnet, L. Chem. Commun. 2002,
20, 2446-2447.
(16) Roux, D.; Chenevier, P.; Pott, T.; Navailles, L.; Regev, O.; Mondain
Monval, O. Curr. Med. Chem., to be published.
Concentrations of nonprotonated hydrazines vary with pH so that
global rate constants cannot be directly compared at different pHs. For
varying pHs, the global rate constant k was corrected into kcorr as
(17) Bonnet, D.; Grandjean, C.; Rousselot-Pailley, P.; Joly, P.; Bourel-Bonnet,
L.; Santraine, V.; Gras-Masse, H.; Melnyk, O. J. Org. Chem. 2003, 68,
7033-7040.
(19) Bourel-Bonnet, L.; Gras-Masse, H.; Melnyk, O. Tetrahedron Lett. 2001,
42, 6851-6853.
(20) Georgheran, K. F.; Stroh, J. G. Bioconjugate Chem. 1992, 3, 138-146.
(18) See Supporting Information for details.
9
16262 J. AM. CHEM. SOC. VOL. 125, NO. 52, 2003

R-Oxohydrazone Ligation on Colloids
A R T I C L E S
described by Sayer et al.:21,22
2.6. Detection of RGDH Ligation Product by Mass Spectroscopy.
LA or control onion vectors (6.5 g/L lipids, ∼85 µM accessible
aldehydes; see calculation below) were incubated for 24 h with RGDH
(170 µM) at pH 5.0 (acetic buffer 10 mM). They were centrifuged (3
h, 4 °C, 26 000 × g), dispersed in buffer and centrifuged again twice
to eliminate excess RGDH. Onion vectors were dispersed in water and
dissolved by adding 2 volumes of ethanol for immediate analysis by
MALDI-TOF MS (Voyager-DE PerSeptive Biosystems, USA) in acid
R-cyano-4-hydroxycinnamic matrix.
2.7. Cell Adhesion Tests. For cell adhesion tests, onion vectors were
prepared as described earlier14 with the following lipid weight composi-
tions: PC/MO 95:5 for control and marker lipid/PC/MO 10:85:5 for
labeled onion vectors, where the marker lipid were LA, LRGD, LRGE,
and DOTAP for LA, RGD, RGE, and cationic vectors, respectively.
Lipid mixes were treated as described above, except that they were
hydrated with a solution of calcein (1 mM, pH 5.6) instead of water in
a 48:52 water/lipids weight ratio. LA vector stock suspensions were
diluted 10 times in buffer pH 4.6 (acetate 10 mM) and incubated for
4 h with 100 µM RGDH or RGEH at room temperature in the dark.
Onion vectors were then centrifuged (30 min, 4 °C, 11 000 × g) and
resuspended in phosphate buffer pH 7.4.
Cell adhesion tests were performed as described earlier.14 Briefly,
endothelial cells of the EAhy-926 cell line23 were harvested using
trypsin, washed thoroughly, mixed with onion vectors, centrifuged
shortly (5 min, 4 °C, 400 × g), and incubated for 4 h at 37 °C in a
CO2 enriched atmosphere. Excess onion vectors were carefully washed
out, cells were harvested with a pipet, and propidium iodide was added
(PI, 0.5 µg/mL) for immediate analysis by flow cytometry (FACScan,
Becton-Dickinson). The relative intrinsic fluorescences q of onion vector
suspensions were measured independently by fluorometry in the
presence of cobalt(II) chloride (Co2+ ions quench the fluorescence of
nonencapsulated calcein, which amounts to 5-40% of total calcein).
The average cell fluorescence of live cells FC was calculated as the
median of the PI-negative data and corrected for intrinsic onion vector
fluorescence as FCcorr ) (FC - FCuntreated)/q, where FCuntreated refers to
the fluorescence of cells treated with no onion vectors.
kcorr K′a + [H+]
k
)
(3)
K′a
where K′ is the protonation constant of the MBTH.
a
2.4. Determination of Thermodynamic Equilibrium Constants.
Stoichiometric mixtures of aldehyde and hydrazine reagents were
prepared at different concentrations, n, and reacted for 24 h at room
temperature. The fluorescence intensity (or alternatively UV absorption
at 250-275 nm for solutions) of the samples was measured. The product
concentration, p, depended on the equilibrium constant, K, and on n as
p ) n + (K-1/2) - (K-1/n)+(K-2/4)). As K was large, p was nearly
x
equal to n. To highlight the slight gap, the relative fluorescence intensity
was plotted versus concentration n as p/pmax - n/nmax and fitted to
equation below:
1
1
nK
1
n +
-
-
+
4K2
n
nmax
2K
x
p/pmax - n/nmax ) M
nmax
-
(4)
1
2K
1
1
+
+
4K2
n
maxK
x
where pmax and nmax refer to the sample of highest concentration, and
M and K are the adjustable parameters (M accounts for experimental
errors on nmax and pmax and was always close to 1). Equilibrium
constants were measured at low pH where hydrazines are protonated.
The apparent constant Kapp was deduced from eq 4, and the real
equilibrium constant was deduced as for rate constants: K ) Kapp(K′
a
+ [H+])/K′.
a
2.5. Acidity and Phase Transfer Constants of MBTH. The
protonation constant of MBTH was measured as the pH of buffer
solutions (i.e., containing 50% of the base and 50% of the protonated
form) of MBTH at different concentrations. nM moles of MBTH
hydrochloride were placed in a vial and diluted to the desired
concentration. The pH was measured under stirring while nM/2 mol of
sodium hydroxide were added. The pH of buffer solutions of concentra-
tions 20, 10, 5, and 2.5 mM were 5.66, 5.67, 5.66, and 5.70,
3. Results
Before preparing and testing grafted vectors as presented in
Figure 1, we studied the reaction of grafting in a more general
case:
respectively, giving the pK′a value with good accuracy: pK′ ) 5.67 at
25 °C.
a
The phase transfer equilibrium constant Kpt of MBTH between water
and lipid colloids was estimated in liposome suspensions. It is defined
as Kpt ) [MBTH]lip/[MBTH]wat where subscripts “lip” and “wat” refer
to concentrations in the lipid phase and the water phase volumes,
respectively. Control liposomes were prepared in buffers at pH 4.6,
7.7, and 3.8. Different dilutions of liposomes (0.15 to 15 g/L lipids)
were incubated with 200 µM MBTH for 24 h at room temperature in
the dark. Liposomes were then removed by ultrafiltration (Microcon
centrifugal filters, Millipore, USA), and the concentration of MBTH in
the bulk was measured by UV absorption. Kpt was deduced by fitting
the following equation to the data:
RR′NsNH2 + R′′COsCHO k,K8
RR′NsNdCHsCOR′′ + H2O (6)
where k is the global rate constant, and K, the global equilibrium
constant of the reaction. The reaction mechanism was first
investigated in water, using the water soluble aldehyde WA
(Figure 2) and N-methylbenzothiazolinone hydrazone (MBTH)
as model reagents. Second, the effect of self-organized colloidal
media on the reaction was studied, in micelles, liposomes, or
onion vectors suspensions. We compared the reaction kinetics
and thermodynamics in two cases: (i) the reaction takes place
mainly in the water bulk (using aldehyde WA) in the presence
of non reactive colloids, and (ii) the reaction takes place mainly
at the surface of colloids where the lipophilic aldehyde (LA) is
sequestered. Third, the results obtained with the model reagent
MBTH were extended to peptidic hydrazines. Finally, the
grafting of peptide ligands onto onion vectors using the
R-oxohydrazone ligation was optimized and tested on cultured
cells.
[MBTH]/[MBTH]wat ) 1 + KptCVj
(5)
where [MBTH] is the initial global concentration of MBTH in the
suspension and CVj is the volume fraction of the lipid phase (considered
negligible compared to the water volume). CVj was estimated as half
of the volume of lipids in the suspension (see below). The measurement
(Kpt ) 600 ( 350) was not very accurate because the separation of
liposomes by ultrafiltration was incomplete. The same experiment using
onion vectors separated by centrifugation gave similar but less accurate
results.
(21) Sayer, J. M.; Peskin, M.; Jencks, W. P. J. Am. Chem. Soc. 1973, 95, 4277-
4287.
(22) Sayer, J. M.; Pinsky, B.; Schonbrunn, A.; Washtien, W. J. Am. Chem. Soc.
1974, 96, 7998-8009.
(23) Edgell, C. J. S.; McDonald, C. C.; Graham, J. B. Proc. Natl. Acad. Sci.
U.S.A. 1983, 80, 3734-3737.
9
J. AM. CHEM. SOC. VOL. 125, NO. 52, 2003 16263

A R T I C L E S
Chenevier et al.
Figure 4. Logarithm of rate constant log(k) of hydrazone formation as a
function of pH, for the reaction of MBTH with WA at 25 °C. (2) Observed
rate constants; ([) rate constants corrected for hydrazine concentration.
Numbers refer to the mechanistic model of Sayer et al. (Figure 3).
Figure 2. Structures of R-oxoaldehyde and hydrazine reagents; amino acids
are represented in the one letter code as bold letters.
Table 1. Rate Constants of Rate Determining Steps for the
Reaction of MBTH with WA at 25 °C, as Measured in Figure 4
i
measured rate
constant
literature data
k1 (M -2 s-1
k2 (M -1 s-1
)
)
4.3 × 105
<5800a,c
100
≈1000a,b
K2k3 (M -2 s-1
K2k4 (M -1 s-1
)
)
8.1 × 105
0.36
<2 × 105 h
0.001d-85h
<1.8 × 106 f
0.3g-7 × 105e
K2K4k5 (M -2 s-1
K2K4k6 (M -2 s-1
)
)
2 × 109
Figure 3. Mechanistic model for imine formation according to Sayer et
al.22 The rate determining steps are indicated by single arrows; noncatalyzed
(2 and 4), acid catalyzed (1, 3, and 5), and base catalyzed (6) steps are
shown as green, red, and blue arrows, respectively.
300
a Literature data: reaction at 25 °C of p-chlorobenzaldehyde with
phenylhydrazine-p-sulfonate. b Literature data: reaction at 25 °C of p-
chlorobenzaldehyde with methoxyamine c Literature data: reaction at 25
°C of p-chlorobenzaldehyde with hydroxylamine d Literature data: reaction
at 25 °C of p-chlorobenzaldehyde with 2-methyl-3-thiosemicarbazide.22
e Literature data: reaction at 25 °C of p-chlorobenzaldehyde with N-
aminoyridinium chloride. f Literature data: reaction at 25 °C of p-
chlorobenzaldehyde with hydrazine. g Literature data: reaction at 25 °C of
p-chlorobenzaldehyde with semicarbazide.21 h Literature data: reaction at
25 °C of p-methoxybenzaldehyde with methoxyamine.24 i Indices of rate
and equilibrium constants (noted k and K, respectively) refer to steps in
the mechanistic model of Sayer et al. (Figure 3).
Although R-oxohydrazone bonds can be detected by UV
absorption, they are difficult to trace in solutions containing
proteins, peptides, or lipid vesicles, which absorb UV light.
Therefore, we turned to a model reaction that could be monitored
as well in peptidic solution as in turbid media. The molecular
address hydrazine was replaced by the fluorescent probe MBTH,
which becomes fluorescent when its hydrazine moiety is
involved in a hydrazone bond. MBTH was reacted with WA or
LA, giving the respective reaction products WP and LP.
3.1. Kinetics in Solution According to Sayer’s Mechanistic
Model. The general mechanism of imine and hydrazone
formation in aqueous solution has been thoroughly described
by Sayer et al.,21,22,24 as summarized in Figure 3. It shows the
addition of a nucleophile amine on a carbonyl compound to
form a hydroxylamine, followed by the dehydration of the
hydroxylamine into an imine. Some or all of the mechanistic
steps can be successively rate determining when changing the
pH of the solution. According to this model, in the case of a
strong nucleophile and a very electrophilic carbonyl, such as
hydrazine and R-oxoaldehyde used in our study, all 6 steps
should be successively rate determining, allowing the determi-
nation of all mechanistic rate constants.
and corrected the rate constant for actual hydrazine concentration
at each pH (Figure 4, diamonds). Depending on pH, the
corrected rate constant was either constant or varied linearly
with proton concentration, highlighting changes in the rate
determining step. As predicted by Sayer’s model, the six steps
depicted in Figure 3 were successively rate determining from
pH 1 to 13; although the plateau corresponding to step 2 was
very small on the pH corrected curve (diamonds Figure 4), it
was evidenced by the discontinuity in the noncorrected curve
around pH 4 (triangles). As shown in Table 1, the values of
each step’s rate constants compared well with previously
published data for the reaction of electrophilic aldehydes and
strong nitrogen nucleophiles.21,22,24-27 However, the rate con-
stants of the acid catalyzed steps (1, 3, 5 in Figure 3) were
unexpectedly high: they were far higher than equivalent rate
constants in reactions involving very strong reagents as hydra-
zine, hydroxylamine, or methoxyamine.
Referring to Sayer’s model, we investigated the reaction of
MBTH with WA in solution as a function of pH. Fluorescence
time series were analyzed according to eq 6 and showed a first-
order kinetics toward each reagent. The global reaction rates k
as a function of pH (Figure 4, triangles) depend both on the
rate determining step and on the protonation of hydrazine into
unreactive hydrazonium. As described by Sayer et al., we
measured the pKa of MBTH (pKa ) 5.67 in water at 25 °C)
As the rate of formation of the hydrazone was optimal in
mild acidic solutions, the pH 4.6 was chosen as the optimal pH
for further experiments.
(25) Sayer, J. M.; Jencks, W. P. J. Am. Chem. Soc. 1977, 99, 464-474.
(26) Hine, J.; Via, F. A.; Gotkis, J. K.; Craig, J. C. J. Am. Chem. Soc. 1970, 92,
5186-5193.
(24) Rosenberg, S.; Silver, S. M.; Sayer, J. M.; Jencks, W. P. J. Am. Chem.
Soc. 1974, 96, 7986-7998.
(27) Hine, J.; Cholod, M. S.; Chess, W. K. J. Am. Chem. Soc. 1973, 95, 4270-
4276.
9
16264 J. AM. CHEM. SOC. VOL. 125, NO. 52, 2003

R-Oxohydrazone Ligation on Colloids
A R T I C L E S
Table 2. Measured Rate Constants k and Equilibrium Constants K for the Reaction of MBTH with WA or LA at 25 °C in Different Aqueous
Mediaa
k (M -1 s-1
at pH 4.6
)
k (M -1 s-1
at pH 7.4
)
K
k )
hyd (s-1
at pH 7.4
medium
(× 106 M-1
)
WA
LA
buffer
micelles 1%
liposomes
micelles 0.25%
micelles 1%
1.33 ( 0.15
2.0 ( 0.3
1.3
7.3 ( 0.5
2.4 ( 0.2
5.1 ( 0.9
0.29 ( 0.07
0.70 ( 0.03
ND
ND
1.5 ( 0.1
1.5 ( 0.3
0.53 ( 0.07
1.9 ( 0.6
ND
ND
97 ( 89
15 ( 5
5 × 10-7
4 × 10-7
ND
ND
1.5 × 10-8
10-7
liposomes or onion vectors
a khyd: rate constant of hydrolysis. ND: not determined.
Table 3. Activation Energies Ea and Arrhenius Prefactors A for
the Reaction of WA or LA with MBTH at pH 4.6, as Determined in
Figure 5
E
a
medium
buffer
micelles 1%
micelles 0.25%
micelles 1%
liposomes
(kJ/mol)
ln(A)
WA
LA
49 ( 5
45 ( 7
57 ( 3
64 ( 6
60 ( 3
20.6 ( 1.7
18.5 ( 2.5
24.8 ( 1.2
26.9 ( 1.5
26.4 ( 1
critical micellar concentration and forms small micelles.28,29 We
therefore considered the measured activation parameters as
mainly driven by the chemical reaction itself.
Contrary to what was expected from its higher rate constant,
the reaction of LA with MBTH had a higher activation energy
than that of WA. Furthermore, within experimental errors,
activation energies depended only on the reagents and not
significantly on the reaction medium. We conclude that the
mechanism of the reaction was not deeply changed in the
presence of colloids and colloids had no catalytic effect. The
gap in activation energy between LA and WA is not understood
but could be due to their structural differences, such as the
protonated lysine in WA which could act as an intramolecular
catalyst.
The reaction rate was mostly increased by changes in the
frequency of reagents encounter, accounted for by the Arrhenius
prefactor A. It is worth noting that A should decrease when one
of the reagents is immobilized on a colloid. Indeed, the
Brownian motion of the immobilized molecule is considerably
reduced compared to a free molecule in solution. Furthermore,
the immobilized molecule can only be approached from a half
space defined by the colloidal surface, whereas free molecules
in solution can be approached from all directions (see Shield et
al. for complete calculations30). Therefore, A can increase only
if both reagents have inhomogeneous concentrations in the
colloidal media and concentrate in the same place.
Figure 5. Reaction rates k of LA or WA with MBTH at pH 4.6 as a function
of the reciprocal temperature. (Open symbols) WA in solution (O) or in
micellar suspension 1% C12E7 (]); (closed symbols) LA in micellar
suspension 0.25% C12E7 (b) or 1% C12E7 (9) and in liposome suspension
(1). Lines are best fits of the Arrhenius equation ln(k) ) ln(A) - Ea/RT.
3.2. Kinetics in Colloidal Suspension. We then compared
the reactivity of MBTH with R-oxoaldehydes in different media
at optimal and physiological pH. Reaction rates with the
lipophilic reagent LA inserted into lipidic colloids (micelles,
liposomes and onion vectors) were compared to reaction rates
with the water soluble reagent WA in buffer or in colloidal
suspension (Table 2). Reaction rates of bond formation were
found to increase up to a factor of 5 in the presence of colloids.
The increase was always small in the case of the reaction of
MBTH with WA but could be high for reactions with LA
embedded in colloids. Considering the Arrhenius model k ) A
exp(-Ea/RT), an increase in k can be due either to a reduced
activation energy, Ea, or to an increase in the prefactor, A. The
former would derive from a catalysis of the reaction, while the
later would emphasize inhomogeneities in the reaction medium.
To distinguish between these hypotheses, the dependence of
rate constants on temperature was investigated.
3.3. Temperature Dependence of Rate Constants. The rate
constants of the reaction of MBTH with LA or WA was measured
as a function of temperature at the optimal pH. They are reported
in an Arrhenius plot in Figure 5. The corresponding activation
energies and prefactors are given in Table 3. Studying the
temperature dependence of reactions in complex media is not
an easy task, as the complex solvent changes with temperature.
As a consequence, activation parameters might contain contribu-
tions from these structural changes. However, our complex
media showed little variation in their physical properties over
the range of temperatures and concentrations used. In particular,
C12E7 at 0.25% and 1% between 5 and 45 °C is well above its
3.4. Partition of Reagents and Products in Different Phases
of Colloidal Media. We investigated possible reagent adsorption
on colloids that could account for their inhomogeneous con-
centrations in colloidal media. LA as well as LP partitioned only
in lipidic colloids because of their two aliphatic chains (100%
of LP fluorescence was found in the colloids after separation
by centrifugation). As a contrary, WA was supposed to be mostly
hydrophilic because of its zwitterionic charge and hydrophilic
tail, whereas the aromatic MBTH is known to be partly
(28) Constantin, D.; Freyssingeas, E.; Palierne, J.-F.; Oswald, P. Langmuir 2003,
19, 2554-2559.
(29) Sharma, K. S.; Patil, S. R.; Rakshit, A. K. Colloids Surf., A 2003, 219,
67-74.
(30) Shield, S. R.; Harris, J. M. J. Phys. Chem. B 2000, 104, 8527-8535.
9
J. AM. CHEM. SOC. VOL. 125, NO. 52, 2003 16265

A R T I C L E S
Chenevier et al.
Figure 8. Kinetics of reaction of MBTH (1 mM) with LA (104 µM) inserted
into liposomes or onion vectors (1.3 g/L lipids) of the same composition,
at pH 4.6 and 25 °C.
Figure 6. Determination of MBTH partition constant between liposomes
and water at pH 7.7 (b), 3.8 (2), and 4.6 ([) at 25 °C; the continuous line
is the best fit to all data of eq 5.
from 0 to 1, a progressive blue shift from 520 to 452 nm was
observed. Fluorescence characteristics similar to those of WP
and LP in colloidal media (i.e., a high fluorescence intensity
and λmax near 488 nm) were recorded for mixtures of dioxane
with the volume fraction near 0.5 and dielectric constant ꢀ ≈
40, corresponding to an environment of intermediate lipophi-
licity. This indicated that the fluorochrome part of WP and LP
adsorbed at the surface of micelles but did not penetrate deeply
into the lipophilic core.
3.5. Effect of Structural Complexity of Colloids: Lipo-
somes versus Onion Vectors. Assuming a homogeneous
repartition in the membranes, LA in liposomes and onion vectors
was shared in two different classes regarding accessibility to
reagents in the bulk water phase: LA molecules displayed on
the surface were directly exposed, whereas those embedded in
inner monolayers were protected from the bulk by at least one
lipid bilayer. To investigate the reactivity of each class, the
reaction yield was compared for reactions of MBTH with LA
onion vectors and liposomes. Both colloids had the same
chemical composition, as liposomes were obtained by high
power sonication of the onion vectors. Although the concentra-
tion and chemical environment of LA were strictly identical,
the reaction yield was expected to depend on the proportions
Figure 7. Maximum emission wavelength (continuous line) and fluores-
cence intensity (dashed line) of the product of reaction of MBTH with WA
as a function of dielectric constant in water/dioxane mixtures at 25 °C;
maximum emission wavelength (triangles) and fluorescence intensity
(circles) of WP (closed symbols) or LP (open symbols) in micellar
suspension (1% C12E7).
hydrophobic (its nonprotonated form is insoluble above 5 mM
in water). However, the partition constants of WA and MBTH
were too low to be measured accurately. Only for MBTH could
the partition constant Kpt between liposomes and water be
roughly estimated to Kpt ) 600 ( 350 (Figure 6). Surprisingly,
no strong dependence on pH was observed. The accuracy of
measurement was limited by the small fraction of liposomes
that could not be filtered out and induced a strong UV light
scattering. The result was also strongly dependent on the
estimation of the volume fraction of the lipid phase, CVj, in which
MBTH dissolved. CVj could not be measured independently; it
was assumed to be half of the total lipid volume in liposomes
(see below).
Adsorption of fluorescent products was also indicated by
changes in emission spectra in the presence of colloids. The
fluorescent product WP had a quantum yield of 10-4 with a
wavelength of maximal emission λmax ) 520 nm in buffer at
pH 4.6 or 7.4 and 25 °C. In micellar solution (1%), its quantum
yield increased to 5 × 10-4 with a blue shift to λmax ) 488 nm.
For comparison, the fluorescence of WP was measured in
mixtures of water and dioxane (Figure 7), whose dielectric
constants can vary from 78 in pure water to 2 in pure dioxane,
mimicking all environments from the aqueous bulk to the inner
core of micelles.31 As the dioxane volume fraction increased
fov and flip of surface exposed LA in onion vectors and liposomes,
respectively. As shown by AFM imaging of the suspensions,
the onion vector preparations were composed of 72%/28% onion
particles/unilamellar vesicles (% in number), whereas the
liposome preparations were reduced to mostly unilamellar
vesicles with 96%/4% unilamellar vesicles/onion particles (data
from J. Thimonier and J. Barbet, not shown). Onion vectors
were made of concentric lipid bilayers of thickness δ, stacked
regularly from the center to the surface with a constant spacing
d.8,9 We assumed that LA was dispersed statistically into all
monolayers, so that fov was equal to the volume ratio of the
outermost monolayer Vom and the total lipid volume of an onion
vector Vtot. We obtain fov ) Vom/Vtot ) d /δ(1 - (1 - δ/2r)3) ≈
3d/2r, where r is the average radius of onion vectors. For our
onion vectors, d was measured to 70 Å by X-ray diffraction,8
and r, to 100 nm by static and dynamic light scattering (data
not shown), so that fov ≈ 0.1. As for liposomes, they were made
of one bilayer containing roughly half of the lipids on each
monolayer, so that flip ≈ 0.5.
Fluorescence monitorings are reported in Figure 8. The yield
of reaction was 5 times lower with onion vectors than with
liposomes, a ratio close to the ratio of surface exposed LA
(31) Lakowicz, J. R. Principles of Fluorescence Spectroscopy; Plenum Press:
1983.
9
16266 J. AM. CHEM. SOC. VOL. 125, NO. 52, 2003

R-Oxohydrazone Ligation on Colloids
A R T I C L E S
Figure 10. Determination of equilibrium constant K for the reaction of
LA in liposomes (1) or WA in solution (O) with MBTH at pH 4.6 and 25
°C. The lines are the best fit of eq 4 to the measured gap to linearity p/pmax
- n/nmax (see Experimental Section).
Figure 9. Apparent reaction rate constant k of LA with MBTH as a function
of pseudophase volume fraction CVj at pH 4.6 and 25 °C: (4) in onion
vector suspension; (b) in liposome suspensions; ([) in micellar suspensions.
The curve corresponds to the best fit of eq 9.
concentration flip/fov. This indicates that only surface exposed
LA had reacted. Then, it seems that MBTH adsorbed but did
not cross bilayers in the time course of our experiments. This
relates well with the low lipophilicity of MBTH and MBTH
hydrazone products as noted above.
3.6. Reaction Rate Dependence on Accessible Lipid
Concentration. The observed rate changes in the reaction of
MBTH with LA were then analyzed in the following pseudophase
model. The reaction was supposed to occur only in the lipidic
pseudophase, where LA was completely sequestered, and its rate
was modulated by the partition of MBTH between the two
phases:
klip of bond formation in the lipidic pseudophase was signifi-
cantly reduced compared to the rate constant measured in
aqueous solution with WA and MBTH. This confirmed that the
reaction of MBTH with LA was locally slower, as the higher
activation energy suggested. Only the high concentration of both
reagents near the colloids due to lipophilic adsorption accounts
for the increase in global reaction rate.
In the case of the reaction of WA with MBTH, the reaction
very probably takes place in both phases simultaneously. The
effect on the reaction kinetics is low though, and the low
accuracy of our measurement did not allow us to identify the
role of each phase in the reaction kinetics.
3.7. Hydrazone Bond Stability. Both the thermodynamic
and kinetic aspects of R-oxohydrazone bond stability were
investigated. We measured first the equilibrium constant K and
then the reaction rate constant k at pH 7.4 to deduce the rate of
hydrolysis khyd at physiological pH as khyd ) k/K. The equilib-
rium constants K were determined by measuring the yield of
reaction at equilibrium as a function of reagent concentrations
(Figure 10). As K was quite high, it was measured at acidic pH
where competition with hydrazine protonation reduced the
apparent equilibrium constants to detectable values. The equi-
librium constants K deduced from these measures are reported
in Table 2. They were found to be very high compared to other
imines or hydrazones. Indeed, reported equilibrium constants
K for the formation of imines and hydrazones in water at 25
°C are at the most in the micromolar range: the most stable
imines are in the millimolar range, whereas the best hydrazones
(p-chlorobenzaldehyde/p-toluenesulfonylhydrazine, K ) 2.4 ×
104 M-1 or p-chlorobenzaldehyde/semicarbazide, K ) 5.5 ×
105 M-1 26,24) and oximes (protonated 2-quinolinecarboxalde-
hyde/hydroxylamine, K ) 1.27 × 105 M-1 32) access the
micromolar range.
MBTHwat{Kpt}MBTHlip
(7)
(8)
MBTHlip + LAlip{klip, Klip}LPlip
where klip and Klip are the reaction rate and equilibrium constants
in the pseudophase and subscripts “wat” and “lip” refer to
solutes in the bulk water phase and the lipidic pseudophase,
respectively. Assuming a rapid partition equilibrium compared
to hydrazone formation, the apparent rate constant k could be
derived as
klipKpt
k )
(9)
1 + KptCVj
where CVj is the volume fraction of pseudophase in the
suspension, C is the global concentration of lipids constituting
the pseudophase, and Vj is the specific volume of those lipids.
As MBTH did not penetrate into complex lipid colloids during
the time scale of the reaction, only the accessible lipids in the
outermost monolayer were considered to take part to the
pseudophase. Vj was estimated from the molecular mass of the
corresponding lipids, as their density is very close to 1 (Vj ≈
0.5 for micelles, Vj ≈ 0.8 for liposomes and onion vectors).
Figure 9 shows the variations of the global rate constant k with
CVj.
Apparent equilibrium constants for LP were higher than those
for WP, as expected in the pseudophase model. Indeed, partition
of the reagents and products into the small volume of the
pseudophase should increase strongly their local concentrations
and favor bond formation. In the model eqs 7 and 8, the global
equilibrium constant K depends on the partition constant Kpt of
MBTH and on the local equilibrium constant Klip in the
Within experimental error, the model fitted well the data and
gave values of Kpt ) 95 ( 50 and klip ) 0.1 ( 0.05 M-1 s-1
Kpt was lower but still in reasonable agreement with the partition
.
constant estimated above to Kpt ) 600 ( 350. The rate constant
(32) Malpica, A.; Calzadilla, M. J. Phys. Org. Chem. 2003, 16, 202-204.
9
J. AM. CHEM. SOC. VOL. 125, NO. 52, 2003 16267

A R T I C L E S
Chenevier et al.
Table 4. Measured Rate Constants k at pH 4.6 and Equilibrium
Constant K for the Reaction of RGDH and MBTH with WA or LA
at 25 °C in Solution or Liposome Suspension
WA/buffer
LA liposomes
k
K
k
K
(M -1s-1
)
(106 M-1
)
(M -1s-1
)
(106 M-1
)
MBTH
RGDH
1.33 ( 0.15
0.53
0.53 ( 0.07
47 ( 20
5.6 ( 0.5a
0.9 ( 0.6a
15 ( 5
NDb
a Determined with competition reaction experiments. b ND: not deter-
mined.
pseudophase as K ) KlipKpt/(1 + KptCVj). Taking into account
the small volume fraction of lipids, CVj, in most of our colloidal
preparations and the relatively high partition constant Kpt, we
expect an increase in global equilibrium constant K of about 2
orders of magnitude between the reaction in water and at the
surface of colloids. This is in fairly good agreement with our
data in Table 2.
3.8. Grafting of Targeting Peptides onto Biological Vec-
tors. We applied the results of our study on the model reaction
to a real application: grafting of RGD peptides as molecular
addresses onto onion vectors. We first compared the reactivity
of the peptide hydrazine RGDH with that of MBTH to check
how far the results obtained with MBTH could be extended to
peptide hydrazines. The reaction rate k at pH 4.6 and equilibrium
constant K of the reaction of RGDH and WA were measured in
solution as previously described for MBTH, using UV absorption
measurement (250-275 nm) instead of fluorometry (Table 4).
Reaction rates on colloids could not be measured by the same
procedure though, as colloidal media absorb most of the UV
light. They were evaluated by an indirect measure: a competi-
tion reaction. LA was reacted simultaneously with MBTH and
RGDH to give products LP and LRGDH at rate constants kMBTH
and kRGDH, respectively:
Figure 11. Kinetics of reaction of MBTH onto LA liposomes, in the
presence or absence of RGDH. The fluorescence intensity of a suspension
of LA containing liposomes (0.3 g/L lipids, accessible LA 10 µM) at pH
4.6 and 25 °C was recorded over time after addition of MBTH 1 mM
(control), MBTH 1 mM + RGDH 1 mM (A), or MBTH 400 µM + RGDH
2 mM (B). the continuous line is the best fit of eq 11 to data B; the dashed
line is the corresponding calculated course of reaction of LA with RGDH.
LA onion vectors incubated with RGDH (2393.6 [M + H+],
2415.5 [M + Na+], exact mass 2393). Unreacted LA was present
in LA onion vectors incubated with or without RGDH (1296.0
[M + Na+], exact mass 1272.5), as expected as only 10% of
LA was accessible to reaction. Trace amounts of RGDH were
detected only on control onion vectors incubated with RGDH
(1138.7 [M + H+], exact mass 1138), showing a weak
adsorption of RGDH onto lipid membranes.
The reactivity of RGDH and MBTH were close as their
reaction rate k in solution were similar. Taking into account its
lower acidity (pK′ ) 6.4533 compared to pK′ ) 5.67 for
a
a
MBTH), RGDH was actually 2 times more reactive than MBTH
at pH 4.6. This higher reactivity was correlated with a 100 times
higher equilibrium constant K for RGDH: the stability of the
RGDH R-oxohydrazone at nanomolar concentrations place it
high above previously described hydrazones, whose stability
do not exceed the micromolar range.26,24
kMBTH
LA + MBTH
8LP + H2O
kRGDH
LA + RGDH
8LRGDH + H2O
(10)
On the other hand, the presence of lipid colloids induced no
such acceleration effect on the reaction of RGDH as that
observed with MBTH. This can be explained since RGDH is
more hydrophilic than MBTH and should not show significant
affinity for lipidic surfaces. However, as pointed out above,
immobilizing the aldehyde reagent onto colloids should decrease
the reaction rate because of its slower motion in the reaction
medium. The similar reaction rates in solution and onto
liposomes are indicative of a very weak adsorption of RGDH
onto lipid bilayers, as confirmed by mass spectroscopy.
3.9. In Vitro Association of Targeted Vectors to Endot-
helial Cells. RGDH was designed to mediate association of
RGDH grafted onion vectors to cells with active integrin
receptors such as endothelial or bone marrow cells. The RGD
ligand included in RGDH was chosen as the hexapeptide
GRGDSP, which is the sequence of fibronectin recognized by
integrins.34 RGDH also included a hydrophilic spacer for grafting
on different kinds of lipid aldehydes.19 The spacer ensured a
minimum distance to the surface of the onion vectors, as it has
Only the apparition of LP was observed (Figure 11), but it was
slowed and limited by the consumption of LA reacting with
RGDH. The reaction kinetics was modeled by the following
equation system:
dp
) kMBTH(n - p)(a - p - p′)
dt
dp′
dt
) kRGDH(n′ - p′)(a - p - p′)
(11)
where p and p′ are the respective instantaneous concentrations
of LP and LRGDH, and n, n′, and a, the respective initial
concentrations of MBTH, RGDH, and accessible LA. Equations
11 can be solved numerically. The calculated p curve was fitted
to fluorescence data to determine the rate constants (continuous
line in Figure 11) from which the reactivity of RGDH was
deduced (dashed line). Rate constants are presented in Table 4.
The formation of LRGDH on onion vectors was evidenced
by mass spectroscopy. After incubation of LA or control onion
vectors with or without RGDH, the vectors were washed and
dissolved in ethanol for immediate analysis by MALDI-TOF
MS (data not shown). LRGDH was detected only in the case of
(33) Bonnet, D.; Ollivier, N.; Gras-Masse, H.; Melnyk, O. J. Org. Chem. 2001,
66, 443-449.
(34) Ruoslahti, E. Annu. ReV. Cell DeV. Biol. 1996, 12, 697-715.
9
16268 J. AM. CHEM. SOC. VOL. 125, NO. 52, 2003

R-Oxohydrazone Ligation on Colloids
A R T I C L E S
4. Discussion
4.1. r-Oxohydrazone: A Kinetically and Thermodynami-
cally Enhanced Hydrazone. A highly hydrolyzable bond such
as hydrazone was not an obvious ligation bond for use in
physiological conditions: although the reaction had the advan-
tages of producing no undesirable subproduct and being
chemoselective, the known stability of hydrazone bonds was at
most in the micromolar range. For controlled modification of
our onion vectors, we required grafted ligands that would stay
in place for the duration of the cell targeting process.
The mechanism deciphered by Sayer et al. for imine and
hydrazone formation predicted a faster and more thermody-
namically favorable reaction in the case of a strong nucleophile
reacting with a highly electrophilic carbonyl.22 This was the
case of our reagents: the R-oxoaldehydes obtained by periodate
oxidation of threonine or serine possess a strongly electron
deprived carbonyl because of the attractive vicinal carbonyl and
the nitrogen in â. On the other hand, peptide hydrazines obtained
by the Bonnet method17 are rendered even more nucleophilic
by the vicinity of the electron rich carbon of the peptide
backbone. In accordance with Sayer’s model, our kinetic data
from the model reaction showed a fast and stepwise mechanism
for hydrazone formation. However, the measured rate constants
were unexpected. The rate constants of uncatalyzed or base
catalyzed steps were within the range of previously reported
data, but the rate constants of acid catalyzed steps (steps 1, 3,
and 5, Figures 3 and 4) were unexpectedly high. We assume
that acid catalysis in those steps was additionally favored by
the vicinal carbonyl which could chelate hydronium ions in close
proximity to the reaction center.
Figure 12. Corrected average fluorescence of live EAhy-926 cells incubated
with onion vectors (0.5 g/L lipids) for 4 h at 37 °C (two independent
experiments); gray bar: neutral control vectors; green bar: cationic vectors;
yellow bars: non grafted LA vectors; red bars: RGD grafted vectors; blue
bars: RGE grafted vectors. Onion vectors grafted by ligation of RGDH
and RGEH (ligation grafted) are compared to vectors labeled with lipid-
ligands LRGD and LRGE as in ref 14 (bulk labeled).
been shown that recognition of grafted RGD peptides by their
cell receptors was inhibited if the ligands are too close to their
support.35 During the synthesis, the hydrazine moiety was
provided by coupling of hydrazinoacetic acid as described by
Bonnet et al.17 As a control, we also prepared an RGE containing
hydrazinoacetyl peptide RGEH, which differs from RGDH only
in the replacement of the aspartate residue by a glutamate.
The vectors were prepared in the optimal conditions deter-
mined above, and their cell targeting efficiency was assessed
in vitro using a previously described protocol.14 Briefly, a
fluorescent dye, calcein, was incorporated into the onion vectors.
Cultured cells from the endothelial cell line EAhy-926 were
incubated over 4 h with onion vectors at 37 °C. Cells were
thoroughly washed, and their fluorescence was measured by
flow cytometry (Figure 12). The fluorescence due to calcein
was corrected by the intrinsic fluorescence of each type of onion
vector, measured independently. As in our previous work,14 the
adhesion ability of RGD onion vectors was compared to those
of neutral and cationic onion vectors. The first showed no
affinity, and the latter, high affinity for the cell surface (Figure
12, controls). Neutral onion vectors containing unreacted LA
behaved mostly like control neutral onion vectors. RGD grafted
onion vectors proved to associate strongly with EAhy-926 cells
compared to those controls. Also, when the RGD ligand
sequence was changed for a RGE sequence, the affinity of
association to cells was lost, showing that interaction of RGD
onion vectors with cells was mediated by molecular recognition
of the grafted ligands by their receptors at the cell surface.
Although the chemoselectivity was not inquired directly in
this study, no competing reaction was observed with the usual
reactive groups present in biological macromolecules. In
particular, the intramolecular primary amine of the lysine residue
in WA could have been a competitor to hydrazines for reaction
with the aldehyde. But this was not observed because hydrazines
are far stronger nucleophiles than amines, as was already shown
for amidation reactions.5 Similarly, the guanidinium group of
the arginine residue in RGDH could have produced a stable
intramolecular bond with the 1,2 dione in the R-oxoaldehyde
moiety. No evidence of a competition of this reaction with the
hydrazone formation was recorded, not surprisingly given the
higher nucleophilicity and lower basicity of hydrazines. Fur-
thermore, a ligation reaction was performed in saturated
saccharose solution, and no competition from the sugar alde-
hydes was observed.
Not only the reaction kinetics was surprising but also the
thermodynamics. Hydrazones of our model reagent MBTH
already showed a stability as high as the best reported carba-
zones. The product of reaction of the peptide hydrazine RGDH
with the peptide aldehyde WA proved even more stable with
an equilibrium constant 2 orders of magnitude higher. Further-
more, even at lower concentrations, bond hydrolysis was very
slow. Indeed, the hydrolysis rate constants estimated at pH 7.4
and 25 °C give for our ligation products half-lives of 23 to 800
days. Even at 37 °C, the half-lives should be long enough for
grafted vectors to touch and attach to their cell target in vivo.
Therefore, the R-oxohydrazone ligation strategy exceeded the
expectations of high reactivity and high stability compared to
Onion vectors grafted using the R-oxohydrazone ligation
strategy were compared to onion vectors containing presynthe-
sized lipid-ligands LRGD and LRGE with the same RGD or
RGE peptide.14 Both types of grafted vectors were very similar
in their behavior toward cells. The presence of R-oxohydrazone
products or unreacted R-oxoaldehyde moieties had neither an
inhibitory effect on binding nor a detectable cell toxicity.
(35) Kantlehner, M.; Finsinger, D.; Meyer, J.; Schaffner, P.; Jonczyk, A.;
Diefenbach, B.; Nies, B.; Kessler, H. Angew. Chem., Int. Ed. 1999, 38,
560-562.
9
J. AM. CHEM. SOC. VOL. 125, NO. 52, 2003 16269

A R T I C L E S
Chenevier et al.
usual hydrazones and finally met the basic requirements for its
use as a ligation for biological applications.
2-fold excess of ligands to increase the rate of reaction but could
have been done in stoichiometric conditions if allowed to react
overnight. Second, as a consequence, problems of chemical
incompatibility or mutual exclusion inside the onion vector
between the ligand and the encapsulated active compound would
be avoided. Third, ligands to graft would be a lot more variable
because hydrazinoacetyl peptides are a lot easier to synthesize
than lipid-ligands. Fourth, the ligation strategy would allow
more complex surface modifications, like the concurrent grafting
of several ligands at the same time. Finally, let us note that no
cytotoxicity was observed, either for R-oxoaldehyde onion
vectors which had not been ligated or by ligation grafted onion
vectors.
4.2. Further Enhancing the Stability Using Lipophilic Self-
Assembly. We intended to use the R-oxohydrazone ligation in
onion vector suspensions, which could affect the kinetics of the
reaction and the stability of the bond. Therefore, effects of self-
organized media on the reaction were still to be investigated.
The hydrazone bond formation proved even faster when
involving the mildly lipophilic MBTH and the lipophilic
aldehyde LA trapped onto colloids. The increase in global
reaction kinetics was not due to a catalysis by colloids (the
activation energy rather increased in colloidal suspension) but
to concentration inhomogeneity in the reaction medium. A
pseudophase model could successfully be applied to our data
not only for micellar media, for which it has mainly been
developed,36 but also for more complex colloids such as
liposomes and onion vectors. In accordance with this model,
the strong anisotropic concentration effects in those suspensions
resulted also in an increase in the apparent stability of the
hydrazone product.
Although this constructive lipophilic self-assembly was absent
or very weak in the case of the ligation of our peptide ligand
RGDH, we plan to take advantage of it in more refined designs.
First, the ligation of large hydrophilic macromolecules onto
lipidic vectors would be slowed by the larger size of the reagent.
It could also be challenged by a displaced hydrophilic/lipophilic
balance, more favorable to anchor solubilization in water thanks
to the macromolecule rather than to fixation of the macromol-
ecule onto the colloid.37 Lipophilic self-assembly could be used
here. For example, a short aliphatic chain could be provided to
the macromolecule together with the hydrazine or R-oxoalde-
hyde moiety prior to grafting. Such modifications are now
accessible even on recombinant proteins.1,6 Autoassociation
would also be of great interest in the formation of structured
peptide loops at the surface of vectors: indeed it has been largely
emphasized that prestructured peptide ligands have stronger
affinities for their receptors.38 Peptide loops could be easily
produced in our ligation strategy by providing the peptide ligand
with an aliphatic chain at the nongrafted end.
5. Conclusion
Our study on a model reaction of the kinetics and thermo-
dynamics of R-oxohydrazone bond formation has shown that
this reaction is an efficient ligation reaction. Submicromolar
stability was achieved as well as half-lives of several weeks
for ligation products in solution. When the ligation was
performed in self-organized media, the kinetics and stability
were enhanced thanks to autoassociation of the reagents. The
ligation of ligands onto colloidal biological vectors can therefore
prove more efficient and favorable than the ligation of soluble
molecules. Those results could be extended to the reagents of
interest, namely targeting peptides grafted as molecular ad-
dresses onto onion vectors. At last, specific cell association tests
proved the biological relevance of this approach.
Our efforts now turn to the use of this ligation strategy for
grafting more complex ligands. The high chemical specificity
of R-oxoaldehyde/hydrazine reactivity allows us to use this
ligation for any kind of protein, glycosylated or not.6 Hydrolysis
of the ligation bond was both unfavorable and very slow; but,
as has been developed for disulfide coupling,2 it could also be
tuned for directed hydrolysis in endosomes, where the acidic
pH catalyzes hydrolysis. Finally, the ligation of macromolecules
could be enhanced in terms of kinetics as well as stability by
taking advantage of the lipophilic autoassociation process we
have observed in our model reaction. Preparation of peptide
loops on lipidic colloids are currently under study.
4.3. r-Oxohydrazone Ligation Proved Efficient on Cells
in Vitro. In our cell association test, RGD onion vectors
obtained with the R-oxohydrazone ligation had excellent target-
ing abilities compared to those of RGD onion vectors using no
ligation. As the two types of vectors proved completely
equivalent for their biological activity, vector modification by
the ligation strategy was preferred as it had many advantages.
First, it provided a means for ligand saving: whereas only 10%
of the lipid-ligand incorporated into the bulk labeled onion
vectors was actually displayed at the surface, all the ligands
grafted onto ligation grafted onion vectors were available for
recognition by cells. Here the ligation was achieved using a
Acknowledgment. We would like to thank O. Melnyk for
his advice; N. Ollivier for providing the WA reagent; K.
Torrecillas for help in the chemical characterization; J. Thimo-
nier and J. Barbet for AFM experiments; H. Drobecq for the
mass spectroscopy analysis; and S. Franc¸ois, B. Delord, and
A. Fe´vrier for their help in cell culture and experiments. This
work was supported by the CNRS and the University of
Bordeaux Sciences and Technology.
Supporting Information Available: Preparation of com-
pounds RGDH, RGEH, LRGD, and LRGE. This material is
available free of charge via the Internet at http://pubs.acs.org.
(36) Lopez-Cornejo, P.; Perez, P.; Garcia, F.; de la Vega, R.; Sanchez, F. J.
Am. Chem. Soc. 2002, 124, 5154-5164.
(37) Silvius, J. R.; Zuckermann, M. J. Biochemistry 1993, 32, 3153-3161.
(38) Thorpe, D. S.; Yeoman, H.; Chan, A. W. E.; Krchnak, V.; Lebl, M.; Felder,
S. Biochem. Biophys. Res. Commun. 1999, 256, 537-541.
JA0370746
9
16270 J. AM. CHEM. SOC. VOL. 125, NO. 52, 2003