Synthesis and evaluation of a near-infrared fluorescent non-peptidic bivalent integrin αvβ3 antagonist for cancer imaging

Article abstract of DOI:10.1021/bc900313d

Computer modeling approaches to identify new inhibitors are essentially a very sophisticated and efficient way to design drugs. In this study, a bivalent nonpeptide intergrin α_vβ₃ antagonist (bivalent IA) has been synthesized on the basis of an in silico rational design approach. A near-infrared (NIR) fluorescent imaging probe has been developed from this bivalent compound. In Vitro binding assays have shown that the bivalent IA (IC₅₀ ) 0.40 ± 0.11 nM) exhibited improved integrin α_vβ₃ affinity in comparison with the monovalent IA (IC50 ) 22.33 ± 4.51 nM), resulting in an over 50-fold improvement in receptor affinity. NIR imaging probe, bivalent-IA-Cy5.5 conjugate, also demonstrated significantly increased binding affinity (IC₅₀ ) 0.13 ± 0.02 nM). Fluorescence microscopy studies showed integrin-mediated endocytosis of bivalent-IA-Cy5.5 in U87 cells which was effectively blocked by nonfluorescent bivalent IA. We also demonstrated tumor accumulation of this NIR imaging probe in U87 mouse xenografts.

Full text of DOI:10.1021/bc900313d

270
Bioconjugate Chem. 2010, 21, 270–278
Synthesis and Evaluation of a Near-Infrared Fluorescent Non-Peptidic
Bivalent Integrin rvꢀ3 Antagonist for Cancer Imaging
Feng Li,† Jiacheng Liu,† Gouri S. Jas,‡ Jiawei Zhang,† Guoting Qin,† Jiong Xing,† Claudia Cotes,† Hong Zhao,†
Xukui Wang,† Laura A. Diaz,† Zheng-Zheng Shi,† Daniel Y. Lee,† King C. P. Li,† and Zheng Li*,†
Department of Radiology, The Methodist Hospital Research Institute, 6565 Fannin Street, B5-022, Houston, Texas 77030, and
Department of Chemistry, Biochemistry, and Institute of Biomedical Studies, Baylor University Sciences Building, Room C322,
Baylor University, 101 Bagby Avenue, Waco, Texas 76706. Received July 14, 2009; Revised Manuscript Received December 22, 2009
Computer modeling approaches to identify new inhibitors are essentially a very sophisticated and efficient way
to design drugs. In this study, a bivalent nonpeptide intergrin Rvꢀ3 antagonist (bivalent IA) has been synthesized
on the basis of an in silico rational design approach. A near-infrared (NIR) fluorescent imaging probe has been
developed from this bivalent compound. In Vitro binding assays have shown that the bivalent IA (IC50 ) 0.40 (
0.11 nM) exhibited improved integrin Rvꢀ3 affinity in comparison with the monovalent IA (IC50 ) 22.33 ( 4.51
nM), resulting in an over 50-fold improvement in receptor affinity. NIR imaging probe, bivalent-IA-Cy5.5 conjugate,
also demonstrated significantly increased binding affinity (IC50 ) 0.13 ( 0.02 nM). Fluorescence microscopy
studies showed integrin-mediated endocytosis of bivalent-IA-Cy5.5 in U87 cells which was effectively blocked
by nonfluorescent bivalent IA. We also demonstrated tumor accumulation of this NIR imaging probe in U87
mouse xenografts.
ure 1, IA) (8). We have also developed a polymerized liposome-
based vascular targeted nanoparticle delivery system using IA
as tumor targeting ligand while carrying a payload for simul-
taneous imaging and therapy (9). In this study, we used IA as
parent compound and constructed a bivalent ligand with a
computer modeling designed linker to enable multivalent ligands
to mimic, compete with, or inhibit natural interactions. Linker
length and rigidity can affect the multivalent ligand-receptor
interaction. In addition, the linkers used should have good
physicochemical properties such as high stability and low
toxicity. Hence, we chose carbon chain as the backbone of
linkers and studied linkers with different lengths by computer
simulation to predict the optimal structure for bivalent ligands
followed by chemical synthesis and in Vitro and in ViVo
validation (Figure 1).
Considering more avid binding could improve the tumor
targeting properties, we further developed a NIR optical imaging
probe based on the bivalent IA for tumor Rvꢀ3 integrin imaging.
The utilization of NIR probes in cancer imaging strategies has
attracted wide attention, because light scattering in tissue
decreases with the reciprocal of the fourth power of wavelength
and hemoglobin has minimal absorbance and autofluorescence
in the NIR window (10, 11). NIR imaging technology is
relatively inexpensive and absent the biological effects of
radioactive probes. In this investigation, our plan was to develop
a cyanine dye (Cy5.5) conjugated NIR fluorescent Rvꢀ3 bivalent
imaging probe (bivalent-IA-Cy5.5) with strong and specific
binding. If successful, this imaging probe has great potential
for early detection of tumors and micrometastatic lesions.
INTRODUCTION
Integrins, a family of cell adhesion molecules, are involved
in a wide range of cell-extracellular matrix (ECM) and
cell-cell interactions (1, 2). Up to now, 18 R and 8 ꢀ subunits
have been described, which assemble into 24 different receptors
(3). Among them, integrin Rvꢀ3 has been intensively studied
due to its important role in tumor angiogenesis and tumor
metastasis. It is highly expressed on activated tumor endothelial
cells and fast-growing tumor cells but not resting endothelial
cells and normal organs; the receptor is therefore suggested as
a molecular target for tumor therapy and imaging. Targeting
Rvꢀ3 by directly inhibiting its function is potentially an effective
strategy for cancer treatment. Cyclic arginine-glycine-aspartic
acid (RGD) peptide, Vitaxin (MedImmune), and cilengitide
(EMD Serono) are currently undergoing clinical trials with
melanoma and glioblastoma multiforme patients (4, 5). Although
a series of monomeric and multivalent RGD peptides have been
reported as an anchor for targeting Rvꢀ3 in molecular imaging,
nonpeptide antagonists were rarely used and few of them have
used rational design to construct the optimal multivalent integrin
Rvꢀ3 ligands (6, 7). In this report, we describe the synthesis
and evaluation of a bivalent nonpeptide small molecule intergrin
Rvꢀ3 antagonist based on an in silico rational design approach.
A near-infrared (NIR) fluorescent imaging probe has been
developed for tumor angiogenesis imaging from this bivalent
compound. We demonstrated the improved receptor binding
affinity, receptor binding specificity of the bivalent IA, and a
promising tumor accumulation of the NIR bivalent IA imaging
probe in mouse xenografts.
Previously, we have developed an integrin Rvꢀ3 targeting
small molecular fluorescent probe derived from a small molecule
Rvꢀ3 antagonist, 4-[2-(3,4,5,6-tetrahydropyrimidine-2-lamino)-
ethyloxy]benzoyl-2-(S)-aminoethylsulfonyl-amino-h-alanine (Fig-
EXPERIMENTAL PROCEDURES
General. All solvents and reagents were purchased from
1
commercial sources and used without further purification. H
NMR and 13C NMR were obtained on a Bruker Ultrashield at
300 and 75 MHz, respectively. Chemical shifts were reported
in ppm (δ) downfield of tetramethylsilane and coupling constants
were given in hertz. The purification of the crude product was
carried out on a semipreparative reversed-phase high-perfor-
* Corresponding author. Tel.: +1 713 441 7962. Fax: +1 713 790
3018. E-mail address: zli@tmhs.org.
† The Methodist Hospital Research Institute.
‡ Baylor University.
10.1021/bc900313d  2010 American Chemical Society
Published on Web 01/26/2010

Bivalent Molecular Imaging Probe by in Silico Design
Bioconjugate Chem., Vol. 21, No. 2, 2010 271
Figure 1. Schematic structure of IA 1 (A); synthetic scheme of bivalent IA 3 (B).
mance liquid chromatography (HPLC) system equipped with a
diode array UV-vis absorbance detector (Agilent 1200 HPLC
system). Mass spectral data were recorded on a ThermoFinnigan
LCQ Fleet using electrospray as the ionization method. U-87
glioblastoma cells were obtained from American Type Culture
Collection and were grown in Dulbecco’s Modified Eagle’s
Medium DMEM (DMEM) supplemented with 10% fetal bovine
serum (FBS) at 37 °C in humidified atmosphere.
Computer Modeling. The AutoDock program (version 3.0.5)
was used for docking calculations (12). The graphical program
SYBYL (version 16.91, Tripos, Inc.) was used for model building
of the ligand structures, to generate missing hydrogen atoms
and assign partial charges for protein targets used in AutoDock
runs, and for energy minimization of initial ligand structures.
Structural figures were generated using WebLab Pro.
rigid docking to the integrin. In our preliminary studies, we only
employed a single conformer for each ligandsthe lowest-energy
structure from the SA run. At this stage, the conformational
analysis was approximate, as a crude solvation model was
employed, with interactions with solvent described by introduc-
ing a distance-dependent dielectric constant.
Simulation of Substrate Binding Using AutoDock. The
auxiliary program AutoGrid was used for generation of affinity
grid fields. The program assigned the center of the grids
automatically. The dimensions of the grid were 60.0 Å × 60.0
Å × 60.0 Å with grid points separated by 0.375 Å. The ligands
were then docked using the Lamarckian Genetic Algorithm
(LGA). The parameters for the LGA configurational search were
identical for all docking jobs. Ligand flexibility was taken into
account by including rotations around bivalent-IA-Cy5.5 single
bonds in the conformational search. The maximum translation
step was kept constant at 0.2 Å for every cycle, but the
maximum quaternion rotation step and dihedral rotation step
had a reduction factor of 0.99 per cycle starting from a
maximum rotation of 5.0°. For the search to be extensive, the
program uses a multiple starting approach in combination with
a time-dependent random number generator. The number of runs
was set to 10. For each run, a population of 50 individuals was
used, with at most 250 000 function evaluations used and run
for at most 27 000 generations. The first top individual was
preserved each generation. The mutation rate was 0.02 and the
crossover rate 0.80. The GA’s selection window was 10
generations; the R parameter was set to 0. After docking
simulations had been carried out, the 10 solutions produced were
grouped into clusters such that the ligand rms deviations within
each cluster were below 0.5 Å. The clusters were also ranked
by the value of the lowest-energy solution within each cluster.
The lowest-energy solution and the solutions with energies
within 5 kcal/mol of this value were considered as possible
binding models.
Bivalent Ligands Synthesis. Bivalent ligand 3 was synthe-
sized in three steps (Figure 1). To a solution of the 2-(((9H-
fluoren-9-yl)methoxy) carbonylamino) octanedioic acid (100 mg,
0.24 mmol) in EtOAc (20 mL), dicyclohexylcarbodiimide (110
mg, 0.53 mmol), and N-hydroxysuccinimide (60 mg, 0.52 mmol)
were added at 0 °C. The mixture was stirred at 0 °C for 8 h
(detected by TLC). Solvent was evaporated under reduced
pressure to get the crude intermediate 2. Compound 2 was
dissolved in DMSO (9 mL) for next step reaction without further
purification. To this DMSO solution, IA (250 mg, 0.50 mmol)
and diisopropylethylamine (0.2 mL, 1 mmol) were added and
the mixture was stirred at ambient temperature for 24 h. After
Protein Coordinates. The structure of extracellular segment
of integrin Rvꢀ3 (PDB entry 1jv2) was used for the docking
simulations of integrin Rvꢀ3. The extracellular segment of the
human integrin structure contains two subunits R (957 residues)
and ꢀ (692 residues). The Rvꢀ3 when complexed with Arg-Gly-
Asp (RGD) reveals that the pentagonal peptide inserts into a
crevice between the propeller and ꢀA domains on the integrin
head (4). The RGD sequence makes the main contact area with
the integrin, and each residue participates extensively in the
interaction, which is completely buried in the protein interior.
For our simulations, we selected subunit ꢀ, and used Glu220
,
Asn215, Asp217, and Pro219 as the target coordinate for binding,
as they are in close contact with RGD. No water molecules
were included. Sybyl was used to add essential hydrogens and
assign partial charges from the Kollman united-atom force field
to the target protein.
Ligand Coordinates. The structures of bivalent-IA-Cy5.5 were
built in SYBYL Biopolymer package and energy minimized
with MMFF94 force field by 100 steps. The orientations of
ligands were overlaid on the integrin Rvꢀ3 in the crystal structure
to allow similar starting coordinates to be used. Prior to docking
runs, all ligand structures were translated arbitrarily 30 Å from
the surface of the Rvꢀ3. Because the ligands are highly flexible,
with multiple rotatable dihedrals present even in the monvalent
system, we adopted a two-stage strategy for protein/ligand
complex structure modeling, involving simulated annealing (SA)
and rigid docking, that partially takes into account ligand
flexibility. In the SA step, a 1 ns MD trajectory was generated
for each ligand at a temperature of 1500 K. Structures saved
every 1 ps underwent energy minimization, and the optimized
structures were sorted by energy. Our long-term goal is to use
5-10 lowest-energy structures of the ligand, thus identified, in

272 Bioconjugate Chem., Vol. 21, No. 2, 2010
Li et al.
100 µL of phosphate-buffered saline (PBS). The cancer cells
were harvested from subconfluent cultures with 0.25% trypsin.
Trypsinization was stopped with DMEM containing 10% FBS,
after which the cells were washed once in serum-free medium
and resuspended in PBS. Trypan blue exclusion was performed
to ensure >90% cell viability. The tumor bearing mice were
subjected to in ViVo imaging studies when the tumors reached
0.4-0.6 cm in diameter.
In Vitro Binding Assay. To determine the receptor-binding
ability, the bivalent ligands were tested for their ability to
competitively inhibit the attachment of the natural ligand
vitronectin to purified human Rvꢀ3 by enzyme-linked immun-
osorbent assay (ELISA). Purified integrin Rvꢀ3 protein (Chemi-
con International, Temecula, CA) was applied to 96-well
polystyrene microtiter plates at 0.1 µg/well. After overnight
incubation at 4 °C, the plates were washed and then blocked
with milk solution (KPL, Inc., Gaithersburg, MD) at room
temperature for 2 h. The blocking buffer was removed, and the
plates were inoculated in quadruplicate with bivalent IAs with
a typical starting concentration of 125 nM. Serial dilutions were
prepared in the 96-well plates using multichannel pipettes.
Biotinylated vitronectin solution (0.1 µg/well) was added to each
well as a standard competitor. The plates were incubated at room
temperature for 3 h, washed, and the bound vitronectin detected
using NeutrAvidin-HRP conjugate at 0.01 µg/well (Pierce,
Rockford, IL) and LumiGlo chemiluminescent substrate system
(KPL, Inc., Gaithersburg, MD). The luminescence was read
using a FLUOstar OPTIMA Microplate Reader (Durham, NC).
The concentration of inhibitor producing 50% inhibition (IC50)
of vitronectin binding to Rvꢀ3 was calculated on the basis of a
curve fitting model using KaleidaGraph 3.5 (Synergy Software,
Reading, PA). c-[RGDfV] (Peptides International, Inc.) was also
tested using the same method described above.
Figure 2. Schematic structure of bivalent-IA-Cy5.5 conjugate.
removing solvent at reduced pressure, 10 mL of water was
added, and the white solid was removed by filtration. The aq
solution was lyophilized, and the residue was recrystallized in
methanol and acetone (1:3) to give pure bivalent IA 3 with 50%
yield as white solid. 1H NMR (300 MHz, D2O): 7.65 (m, 4H),
6.93 (m, 4H), 4.16 (m, 7H), 3.84 (m, 4H), 3.30 (m, 10H), 3.21
(m, 10H), 2.05 (t, 2H, J ) 7.0 Hz), 1.80 (m, 6H), 1.43 (m,
2H), 1.18 (m, 4H). MS (electrospray): m/z 1066.3 (100,
[M+H]+, calculated 1065.4); 533.8 ([M+2H]2+. calculated
533.7).
Fluorescence Microscopic Studies. To demonstrate that the
bivalent-IA-Cy5.5 imaging probe can act as a specific ligand
for Rvꢀ3 integrin receptor in living cells, and to demonstrate
the binding and subcellular localization of the bivalent probe,
U-87 MG (ATCC HTB-14), a human glioblastoma cell line
known to overexpress integrin Rvꢀ3, was used in a probe binding
assay (4). For fluorescence microscopy studies, 1 × 104 U87
cells were cultured on 35 mm MatTek glass-bottomed culture
dishes. After 24 h, cells were washed with PBS and then
incubated with bivalent-IA-Cy5.5 (200 nM) at 37 °C for 45
min. Integrin specificity of the probe binding was verified by
incubating U87 cells with and without a blocking dose of the
nonfluorescent bivalent IA (20 µmol/L). After the incubation
period, the cells were washed three times with ice-cold PBS.
The fluorescence signal of the cells was recorded using
FluoView 1000 laser scanning confocal microscope (Olympus)
under a 40× oil immersion objective, with laser excition at 675
nm (Cy5.5). A differential interference contrast image was also
taken so that the origin of fluorescent signals could be confirmed.
In ViWo Fluorescence Imaging of Tumors. ∼3 × 106 U-87
cells were implanted bilaterally in the posterior flanks of
immunocompromised (nu/nu) mice and grown to ∼4-6 mm
in diameter. Imaging was carried out with a group of 5 mice.
After palpable masses were detected, animals were manually
restrained and imaging probes were intravenously administered
(lateral tail vein) using a 30 G needle via aseptic technique.
Free Cy5.5 was prepared by dissolving the same Cy5.5NHS
ester used for labeling the bivalent IA (GE Healthcare) in
phosphate buffered saline (PBS), pH 7.4, overnight at 4 °C.
Stock concentration of free Cy5.5 was measured optically using
the molar exctinction coefficient of 250 000 M-1 cm-1 using a
UV-visible spectrophotometer using the maximum absorbance
wavelength (675 nm) as recommended by the supplier (GE
Healthcare). The bivalent-IA-Cy5.5 was weighed and confirmed
Bivalent-IA-Cy5.5 Conjugate Synthesis. (Figure 2) Bivalent
IA 3 (1.5 mg, 1.4 µmol) and 1.2 equiv of Cy5.5-NHS (2.0 mg,
1.7 µmol) were dissolved in 1 mL DMSO, and then 0.1 mL
triethylamine was added. The reaction mixture was stirred in
the dark at ambient temperature overnight and then quenched
by adding 200 µL of 5% acetic acid (HOAc). The crude product
was purified by a semipreparative reversed-phase HPLC em-
ploying a PhenomenexLuna C-18 column (250 mm × 10 mm)
with the mobile phase starting from 95% solvent A (0.05%
ammonia acetate in water) and 5% solvent B (acetonitrile) to
40% solvent A and 60% solvent B at 30 min, and the flow rate
is 4 mL/min. Fractions containing bivalent-IA-Cy5.5 conjugates
were collected, lyophilized, redissolved in saline at a concentra-
tion of 1 mg/mL, and stored in the dark at -20 °C until use.
The analytical HPLC method was performed with the same
gradient system, but with a Phenomenex C-18 column (250 mm
× 4.6 mm) and flow rate of 1 mL/min. The purity of Cy5.5-
labeled bivalent IA was over 98% from analytical HPLC
analysis. The yield of bivalent-IA-Cy5.5 conjugate was over
70%. The purified bivalent-IA-Cy5.5 conjugate was character-
ized by mass spectroscopy (MS). MS (electrospray): m/z 1965.7
([M+H]+, calculated 1965.2); 983.1 (100, [M+2H]2+, calculated
983.6). The absorption and fluorescence emission characteristics
of bivalent IA-Cy5.5 conjugates were identical to those of free
Cy5.5, as apparent from the spectra measured in H2O (data not
shown).
U-87 MG Xenograft Model. Animal procedures were
performed according to a protocol approved by The Methodist
Hospital Research Institute Animal Care and Use Committee
(IACUC). Female athymic nude mice (nu/nu), obtained from
Charles River Laboratories, Inc. (Cambridge, MA) at 4-6 weeks
of age, were injected subcutaneously in the right rear thigh with
1 × 107 U-87 MG cells (see general procedures) suspended in

Bivalent Molecular Imaging Probe by in Silico Design
Bioconjugate Chem., Vol. 21, No. 2, 2010 273
Table 1. Conformational Energy of Bivalent IA and NIR Imaging Probe
by spectrophotometer measurements. Appropriate dilutions of
each probe were made in PBS, pH 7.4, and filter sterilized (0.2
µm) immediately prior to use. All animals weighing between
18 and 20 g received 10 nmol of imaging probe (total volume
of 100 uL). Fluorescence imaging was performed with a small
animal dedicated optical imaging system (Xenogen/Caliper
IVIS-200, Mountain View, CA) under continuous 1.5-2.0%
isoflurane delivered through the integrated anesthesia system.
At 2, 4, 7, 24, and 48 h postinjection, images were acquired
with the surface of the tissue of interest (tumor) approximating
a perpendicular viewing angle relative to the camera line of
sight; this viewing angle was also used for normal tissue
selection. For determining tumor contrast, mean fluorescence
intensities of the tumor area (T) defined as radiance (photons/
s/cm2/sr) at the right rear leg of the animal and of the area (N)
at the right flank (normal tissue) were calculated by region-of-
interest (ROI) analysis using the corresponding photograph of
each image acquisition data set to identify the tumor region.
Preinjection imaging was also performed to determine baseline
autofluorescence. For competitive inhibition studies, mice (n
) 3) were coinjected with 500 nmol bivalent IA and 10 nmol
of imaging probe (bivalent-IA-Cy5.5) following the same
procedure described above. Free Cy5.5 studies were performed
in the same manner following the intravenous injection of 10
nmol. For ex vivo biodistribution imaging, two tumor-bearing
mice were imaged 24 h post injection of bivalent-IA-Cy5.5 and
thereafter euthanized per IACUC protocol (CO2 and cervical
dislocation). Tumor and major internal organs (heart, lung, liver,
spleen, gastrointestinal tract, and kidneys) were harvested,
weighed, and imaged against a black paper background. All
image acquisition parameters were as follows: epi-illumination;
Cy5.5 Em/Ex; Bin: (HR)4, FOV: 12.9; f2; 0.2 s.
orientation of the ligand relative to the protein target. The
solutions were clustered on the basis of rms (root-mean-square)
deviations in ligand atomic positions, with structures with rmsd
of less than 0.5 Å grouped into a cluster. The total number of
generated low-energy clusters measures the specificity of
binding (12–15). A small number of clusters indicates that the
ligand has only a few possible binding modes and interacts with
a specific site (or sites) on the target protein. On the other hand,
a large number of clusters implies existence of a wide range of
binding modes and lack of specific ligand-target interactions.
The second step in sorting solutions involves identification of
the solution of lowest binding energy within each cluster and
ranking the different clusters according to this energy value.
The solution with the lowest energy in the top-ranked cluster
and all solutions with energies higher by up to 5.0 kcal/mol
were considered as possible binding modes for ligand to target.
Docking of Bivalent-IA-Cy5.5 to the Integrin rvꢀ3. A
summary of the AutoDock results is presented in Table 1; the
docked structures of the first and second clusters are shown in
Figure 3. The docking of bivalent-IA-Cy5.5 to the integrin Rvꢀ3
produced 3 clusters out of 10 runs. There were 5 solutions in
the first cluster with an average docking energy of -34.04 kcal/
mol. The binding site defined by this cluster is in the cleft
between the RGD binding domain in integrin Rvꢀ3. It is observed
that bivalent-IA-Cy5.5 stays at that site in the presence of the
whole protein. This cluster presents the best model for a possible
bivalent-IA-Cy5.5 interaction to the integrin Rvꢀ3. A comparison
of the AutoDock results suggests that this binding conformation
corresponding to the linker length of the bivalent-IA-Cy5.5 is
a better substrate of integrin Rvꢀ3. The binding of bivalent-IA-
Data Processing and Statistics. All data are given as means
( SD of n independent measurements. Statistical significance
was assigned for P values less than the boundaries of tumors.
Quantitative data were acquired with the software provided from
the instrument’s manufacturer (LiVing Image, v 3.1). Statistics
software from R Development Core Team (2009). (R: A
language and environment for statistical computing. R Founda-
tion for Statistical Computing, Vienna, Austria. ISBN 3-900051-
07-0, URL http://www.R-project.org).
RESULTS
Computer Modeling. From calculations of binding energies
and ligand spatial distributions, the prediction of favored
protein-ligand binding modes, interaction strengths, and binding
specificity can be obtained with AutoDock simulations. Ten (10)
LGA (Lamarckian genetic algorithm) docking runs were
performed for each protein-ligand pair, with each run producing
one possible binding mode or solution. The solutions were first
sorted in terms of the binding mode, i.e., the position and
Figure 3. The lowest-energy conformation of bivalent-IA-Cy5.5
conjugate (A). Interaction of bivalent-IA-Cy5.5 with the Integrin (B).
The lowest energy conformation is obtained from the analysis of
molecular dynamics trajectories. Integrin structure is obtained from the
open form of integrin crystal structure (1JV2).

274 Bioconjugate Chem., Vol. 21, No. 2, 2010
Li et al.
antagonists and biotinylated human vitronectin for the im-
mobilized integrin Rvꢀ3. The IC50 value for parent compound
IA, bivalent IA, bivalent-IA-Cy5.5 conjugate, and c-[RGDfV]
were 22.33 ( 4.51, 0.40 ( 0.11, 0.13 ( 0.02, and 4.80 ( 3.01
nM, respectively (Figure 4). Cy5.5 conjugation did not decrease
the receptor binding affinity of the resulting NIR fluorescent
bivalent IA.
Fluorescence Microscopy. Bivalent-IA-Cy5.5 conjugate was
incubated with U87 cells known to overexpress integrin Rvꢀ3.
Images were taken with a confocal fluorescent microscope.
Receptor-mediated endocytosis was observed in the cells. The
binding of bivalent-IA-Cy5.5 to U87 cells was effectively
blocked by coincubation with nonfluorescent bivalent IA (Figure
5).
In ViWo Imaging of Integrin Overexpression in
Xenograft U-87 Tumors. Whole animal fluorescence imaging
of subcutaneously implanted U-87 xenograft tumors demon-
strated increased signal in tumor sites at 7, 24, and 48 h
postconstrast injection (Figure 6). Imaging at earlier time points
(2 and 4 h) showed intense systemic fluorescence signal due to
relatively slow clearance of the compound. However, at 7 h,
there is sufficiently high tumor-to-normal tissue (T/N) contrast
to clearly discern the lesion; and by 24 and 48 h, the most
intense signal is evident at the tumor site. Region of interest
(ROI) analysis was performed by determining the boundaries
of superficial tumors based on photographic images; this region
was then measured for fluorescence signal (radiance units,
photons/s/cm2/steradian). Normal tissue was selected by copying
the ROI of each sample to a site in the anterior flank. As plotted
in Figure 7, T/N ratios reach a plateau at 24 h with an average
maximum of ∼1.84. For the 7, 24, and 48 h time points, the
fluorescence signals of the tumor and normal tissues were
statistically significantly different (two sample, one side t test,
p-value 0.006, 0.015, and 0.007, respectively, n ) 5).
Figure 4. Enzyme-linked immunosorbent assay (ELISA) of IA (IC50:
22.33 ( 4.51 nM), bivalent IA (IC50: 0.40 ( 0.11 nM), bivalent-IA-
Cy5.5 conjugate (IC50: 0.13 ( 0.02 nM), and c-[RGDfV] (IC50: 4.80
( 3.01 nM). All of the points were done in triplicate.
Cy5.5 to the protein is stronger (lower energy) and more specific
(small number of clusters) compared to other conformations with
different linker length. The microscopic reason for these effects
appears to be a lack of fit between different Cy5.5 conformation
and the inhibitor/substrate binding site. The docking results are
approximate. The scoring is based on an empirical energy
function, solvation effects treated with a highly simplified model,
and only ligand flexibility taken into account, with the protein
structure kept fixed (12–15). Thus, the AutoDock results should
only be considered as qualitative. The calculated binding energy
results are in qualitative agreement with the observed inhibitory
effects. Additionally, the simulations suggest that bivalent-IA-
Cy5.5 with this conformation has a higher specificity and better
fit into the known active site than other conformations.
To ascertain the specificity of this tumor tissue signal
localization, competitive inhibition studies were carried out using
the nonfluorescent (without Cy5.5) bivalent IA. Using the same
tumor model, a 10-fold excess of bivalent IA was coadminis-
tered with 10 nmol of bivalent-IA-Cy5.5 and were imaged at
In Vitro Binding Affinity. Enzyme-linked immunosorbent
assay (ELISA) was used to measure the competitive binding of
Figure 5. Specific binding of the bivalent-IA-Cy5.5 with U-87 cells. The left (A,D), middle (B,E), and right panels (C,F) represent fluorescence,
brighten, and fusion images, respectively, taken from the same field of view. Confocal laser-scanning microscopy images of U-87 glioblastoma
incubated with bivalent-IA-Cy5.5 demonstrated endocytosis of the imaging probes (A-C). U-87 cells blocking experiment with nonfluorescent
bivalent IA demonstrated a complete blockage of endocytosis of the imaging probes (D-F).

Bivalent Molecular Imaging Probe by in Silico Design
Bioconjugate Chem., Vol. 21, No. 2, 2010 275
Figure 6. In vivo fluorescence imaging of live U-87 xenograft tumor-bearing mice. Photograph images (top panel) of representative mouse subject
at selected time points. Photon scale is shown in the right; corresponding near-infrared fluorescence imaging (bottom panel). Radiance scale is
shown in the color bar to the right (p/s/cm2/sr). Subject was anesthetized with 1.5-2.0% isoflurane through a nose-cone. Arrow indicates tumor.
Figure 7. Region of interest (ROI) analysis of fluorescence signal
(radiance) from tumor relative to normal tissue (T/N) vs time course
following bivalent-IA-Cy5.5 administration (in hours). Error bars at
each time point represents standard deviation from n ) 5 subjects.
Figure 8. Competitive inhibition of bivalent-IA-Cy5.5 conjugate with
2, 4, 7, 24, and 48 h. ROI analysis was again performed as
described above. When compared between blocking and non-
nonfluorescent bivalent IA. Significant blocking of tumor fluorescence
signal (y-axis) was observed at all time points (x-axis). Units of radiance
blocking controls, fluorescence signals within tumors were
reduced at all time points measured but were most pronounced
in the earlier time points (2, 4, and 7 h), perhaps due to differing
clearance rates of the two compounds (Figure 8). Nonetheless,
the observed blocking effect was significant at all time points
(two samples, one side t test, p ) 0.001 (2 h); 0.002 (4 h);
0.003 (7 h); 0.012 (24 h); and 0.019 (48 h); n ) 6 (blocking);
n ) 4 (nonblocking control). As a negative control, two tumor-
bearing animals were intravenously injected with 10 nmol of
the free Cy5.5 and imaged at 24 h post-injection. As shown in
Figure 9, there is an overall reduction in systemic fluorescence
signal from the entire animal. The relative amount of fluores-
cence signal from the tumor appears to be slightly higher than
the background but is lower than that achieved with the bivalent-
IA-Cy5.5 (Figure 9, compare left and right panels). This result
is best explained by a higher clearance rate of the free
fluorophore and a certain degree of nonspecific binding to
tumors as observed by other investigators (16). Collectively,
these data support the specificity of the bivalent-IA-Cy5.5 for
integrin Rvꢀ3 associated with U-87 tumors.
(p/s/cm2/sr). Error bars indicate SD from SD (n ) 6 blocked, n ) 4
nonblocked).
nous administration of the bivalent-IA-Cy5.5, two tumor-
bearing animals were imaged as described above, imaged at
24 h postinjection and immediately thereafter euthanized for
harvest of internal organs. Tumors and the major organssheart/
lung (en bloc), liver, spleen, gastrointestinal tract, and
bilateral kidneysswere resected and placed on a black
background and imaged under the same imaging acquisition
protocol (Figure 10). Quantitative analysis was performed
by measuring the total photon flux of each selected tissue
and divided by the mass (Figure 11). The highest signal per
mass of tissue was observed for the tumor, followed next by
the heart and lung (en bloc), and kidneys. These data indicate
that the bivalent-IA-Cy5.5 is specific for the tumor tissue,
likely through binding to its cognate integrin receptor. Given
the moderate level of fluorescence signal in the heart and
lung, this compound is likely to have a relatively slow blood
pool clearance rate. Fluorescence signal from the kidneys
suggests mainly renal clearance of the agent, and minimal
liver signal suggest low hepatic metabolism. Detailed bio-
Given the intrinsic limitation in tissue penetration of
excitation and emission photons from fluorophores, ex vivo
imaging was performed to assess the fluorescence signal of
the tumor relative to internal organs. Following the intrave-

276 Bioconjugate Chem., Vol. 21, No. 2, 2010
Li et al.
Figure 11. Biodistribution of bivalent IA-Cy5.5 at 24 h post-injection
averaged from 2 subjects. Y-axis is total photon flux per gram of tissue.
gliomas present some of the greatest challenges in the manage-
ment of cancer patients worldwide, despite notable recent
achievements in oncology. Even with aggressive surgical
resections using state-of-the-art preoperative and intraoperative
neuroimaging, along with recent advances in radiotherapy and
chemotherapy, the prognosis for GBM patients remains dismal.
Mean survival after diagnosis is about 1 year (18). Preclinical
data indicate that angiogenesis is essential for the proliferation
and survival of malignant glioma cells, which suggests that
inhibition of angiogenesis might be an effective therapeutic
strategy.
Figure 9. In ViVo fluorescence imaging of live U-87 xenograft tumor-
bearing mice at 24 h postintravenously administered unconjugated Cy5.5
(left panel) vs bivalent IA-Cy5.5 (right panel). Photograph images (top
panel) of representative mouse subject. Photon scale is shown in the
right; corresponding near-infrared fluorescence imaging (bottom panel).
Radiance scale is shown in the color bar to the right (p/s/cm2/sr). Mice
were anesthetized with 1.5-2.0% isoflurane through a nose-cone. Black
paper was used to cover the injection sites in the lateral tail veins.
Arrows indicates tumors.
distribution, pharmacokinetic, and toxicity studies will be
reported separately.
NIRF imaging has been proven to be a very powerful tool
for noninvasive imaging of various diseases in preclinical
models especially in rodents (10). In an effort to leverage this
robust modality, we sought to apply NIRF imaging techniques
for studying an important molecular target, integrin Rvꢀ3,
expressed by glioblastoma cells. Our ultimate goal was to
develop imaging and therapeutic agents targeted to integrin Rvꢀ3
for early detection, treatment, and monitoring of glioblastoma
as well as other malignancies. To achieve this goal, we set out
to design and develop a highly specific and sensitive NIRF agent
for this clinically important target.
DISCUSSION
Glioblastoma multiforme (GBM) is the most common
primary brain tumor, as well as the deadliest (17). Malignant
Computer-assisted approaches to identify new inhibitors via
pharmacophore, molecular modeling, docking, and structural
interaction fingerprints is rapidly becoming an integral part of
rational approaches to drug design. In silico assessment of the
free energy and the binding affinity of receptor-inhibitors prior
to synthesis permits sophisticated and efficient methods to design
drugs. Hood et al. demonstrated that the integrin antagonist,
4-[2-(3,4,5,6-tetrahydropyrimidine-2-lamino)ethyloxy]benzoyl-
2-(S)-aminoethylsulfonyl-amino-h-alanine (IA) can be used as
a targeting agent for gene delivery to tumor neovasculature (2).
A rationally designed bivalent IA by computer modeling could
theoretically have better binding affinity and thus greater avidly
to target cells expressing this receptor. In an effort to develop
the bivalent IA probe for tumor angiogenesis imaging, an
extensive simulation with varying linker lengths and composition
was carried out. Our results (shown in Table 1) demonstrated
that bivalent IAs with linkers containing 5 to 8 carbons (n )
2-5) exhibited descending conformational energy, while the
linker shorter than 5 carbons (n < 2) had a significantly higher
energy. On the basis of these simulation results, we chose the
8-carbon linker (n ) 5), which has the lowest conformational
energy to construct the first bivalent IA. Among various
commercially available linkers, 2-aminododecanedioic acid,
which carries two carboxylic acid groups at each end with a
total of eight carbons (n ) 5), was chosen as the best linker for
the prototype construct. This analysis predicted that the bivalent
IA with an 8-carbon linker (n ) 5) containing the NIR
fluorophore, Cy5.5 dye, would not sterically impact the binding
to the integrin protein.
Figure 10. Representative ex ViVo NIR imaging of major internal organs
harvested from a xenograft tumor-bearing mice at 24 h post i.v. injection
of bivalent IA-Cy5.5. Top panel: photograph (photon scale is shown
to the right). Bottom: fluorescence image (p/s/cm2/sr). Harvested organs
are as follows: (k) kidneys; (li) liver; (h/l) heart and lungs; (gi) small
and large intestines; (t) U-87 tumor; (sp) spleen.

Bivalent Molecular Imaging Probe by in Silico Design
Bioconjugate Chem., Vol. 21, No. 2, 2010 277
In Vitro binding assays have shown that the bivalent IA
(IC50 ) 0.40 ( 0.11 nM) exhibited significantly improved
integrin Rvꢀ3 affinity when compared to the parent compound
IA (IC50 ) 22.33 ( 4.51 nM), resulting in a 50-fold
improvement in receptor affinity (IC50) over that of the parent
compound IA and a 10-fold improvement over c-[RGDfV]
(IC50 ) 4.80 ( 3.01 nM). NIR imaging probe, bivalent-IA-
Cy5.5 conjugate, also demonstrated significantly increased
binding affinity (IC50 ) 0.13 ( 0.02 nM). Fluorescence
microscopy studies demonstrated integrin-mediated endocy-
tosis of the bivalent-IA-Cy5.5 conjugate in U-87 cells, which
was effectively blocked by nonfluorescent bivalent IA. This
result provides strong supporting evidence that bivalent-IA-
Cy5.5 binds specifically to the Rvꢀ3 integrin receptor
expressed on the tumor cell surface.
Systemically administered bivalent-IA-Cy5.5 in tumor-
bearing mice resulted in modest accumulation at the tumor
site with improved tumor/normal tissue signal over time (up
to 48 h). This effect is best explained by the relatively slow
clearance of the probe from the circulation as evident from
the high fluorescence observed throughout the test animals
at earlier time points (up to 24 h). Consistent with prior
studies using a monomeric integrin peptidomimetic conju-
gated to a chelate for radioisotope imaging, tumor accumula-
tion was time-dependent and required washout from normal
tissue for optimal contrast (8). Although the tissue penetration
of excitation and emission photons of near-infrared light is
superior to wavelengths of the visible spectrum, a substantial
percentage of light is attenuated as a function of tissue depth.
Hence, fluorescence measurements likely underestimate the
true accumulation of integrin-specific probes at the target
tissue.
The experimental results presented here demonstrate the
ability to noninvasively image integrin Rvꢀ3 overexpression
in live whole animals using a bivalent small molecule with
improved receptor binding properties generated by in silico
design. A major impetus for our experimental design was to
demonstrate proof of concept in generating a new class of
targeted imaging agents by leveraging known structure
activity relationships and in silico modeling to further
improve receptor binding properties of existing molecules.
Prior multivalency studies have clearly shown the utility of
expanding the repertoire of drugs by reconfiguring molecules
as multimers (19, 20). We propose that such approaches will
have significant impact in the drug discovery process by
providing a mechanism to alter the pharmacologic properties
of existing as well as novel drugs.
Baylor Supercomputing Center and internal funding. We ap-
preciate Dr. Shi Ke for discussion and advice on biological
study.
LITERATURE CITED
(1) Brooks, P. C., Clark, R. A., and Cheresh, D. A. (1994)
Requirement of vascular integrin alpha v beta 3 for angiogenesis.
Science 264, 569–71.
(2) Hood, J. D., Bednarski, M., Frausto, R., Guccione, S., Reisfeld,
R. A., Xiang, R., and Cheresh, D. A. (2002) Tumor regression
by targeted gene delivery to the neovasculature. Science 296,
2404–7.
(3) Takagi, J., and Springer, T. A. (2002) Integrin activation and
structural rearrangement. Immunol. ReV. 186, 141–63.
(4) Delbaldo, C., Raymond, E., Vera, K., Hammershaimb, L.,
Kaucic, K., Lozahic, S., Marty, M., and Faivre, S. (2008) Phase
I and pharmacokinetic study of etaracizumab (Abegrin), a
humanized monoclonal antibody against alphavbeta3 integrin
receptor, in patients with advanced solid tumors. InVest. New
Drugs 26, 35–43.
(5) Schottelius, M., Laufer, B., Kessler, H., and Wester, H. J. (2009)
Ligands for mapping Rvꢀ3-integrin expression in vivo. Acc Chem
Res 42, 969–80.
(6) Chen, X., Conti, P. S., and Moats, R. A. (2004) In vivo near-
infrared fluorescence imaging of integrin alphavbeta3 in brain
tumor xenografts. Cancer Res. 64, 8009–14.
(7) Cheng, Z., Wu, Y., Xiong, Z., Gambhir, S. S., and Chen, X.
(2005) Near-infrared fluorescent RGD peptides for optical
imaging of integrin alphavbeta3 expression in living mice.
Bioconjugate Chem. 16, 1433–41.
(8) Burnett, C. A., Xie, J., Quijano, J., Shen, Z., Hunter, F., Bur,
M., Li, K. C., and Danthi, S. N. (2005) Synthesis, in vitro, and
in vivo characterization of an integrin alpha(v)beta(3)-targeted
molecular probe for optical imaging of tumor. Bioorg. Med.
Chem. 13, 3763–71.
(9) Xie, J., Shen, Z., Li, K. C., and Danthi, N. (2007) Tumor
angiogenic endothelial cell targeting by a novel integrin-targeted
nanoparticle. Int. J. Nanomed. 2, 479–85.
(10) Ntziachristos, V., Bremer, C., and Weissleder, R. (2003)
Fluorescence imaging with near-infrared light: new technological
advances that enable in vivo molecular imaging. Eur. Radiol.
13, 195–208.
(11) Sevick-Muraca, E. M., Houston, J. P., and Gurfinkel, M.
(2002) Fluorescence-enhanced, near infrared diagnostic imaging
with contrast agents. Curr. Opin. Chem. Biol. 6, 642–50.
(12) Morris, G. M., Goodsell, D. S., Halliday, R. S., Huey, R.,
Hart, W. E., Belew, R. K., and Olson, A. J. (1998) Automated
docking using a Lamarckian genetic algorithm and empirical
binding free energy function. J. Comput. Chem. 19, 1639–
1662.
(13) Goodsell, D. S., and Olson, A. J. (1990) Automated docking
of substrates to proteins by simulated annealing. Proteins 8, 195–
202.
(14) Morris, G. M., Goodsell, D. S., Huey, R., and Olson, A. J.
(1996) Distributed automated docking of flexible ligands to
proteins: parallel applications of AutoDock 2.4. J. Comput. Aided
Mol. Des. 10, 293–304.
(15) Xiong, J. P., Stehle, T., Zhang, R., Joachimiak, A., Frech,
M., Goodman, S. L., and Arnaout, M. A. (2002) Crystal
structure of the extracellular segment of integrin alpha Vbeta3
in complex with an Arg-Gly-Asp ligand. Science 296, 151–5.
(16) Cheng, Z., Levi, J., Xiong, Z., Gheysens, O., Keren, S.,
Chen, X., and Gambhir, S. S. (2006) Near-infrared fluorescent
deoxyglucose analogue for tumor optical imaging in cell
culture and living mice. Bioconjugate Chem. 17, 662–9.
(17) Parsons, D. W., Jones, S., Zhang, X., Lin, J. C., Leary, R. J.,
Angenendt, P., Mankoo, P., Carter, H., Siu, I. M., Gallia, G. L.,
Olivi, A., McLendon, R., Rasheed, B. A., Keir, S., Nikolskaya,
T., Nikolsky, Y., Busam, D. A., Tekleab, H., Diaz, L. A., Jr.,
Hartigan, J., Smith, D. R., Strausberg, R. L., Marie, S. K., Shinjo,
CONCLUSION
We have successfully synthesized and evaluated a novel
near-infrared fluorescent nonpeptidic bivalent integrin Rvꢀ3
antagonist for tumor imaging with improved binding avidity
relative to the monovalent ligand. The design of this bivalent
compound was facilitated by in silico modeling which guided
synthetic strategy in a practical application of optimizing a
biologically relevant imaging probe. We suggest that this
approach represents a general paradigm to improve receptor
binding affinities of extant compounds through rationally
designed multivalent configurations. More research will be
carried out to rationally design bivalent ligands with different
types of linkers using computer modeling followed by in Vitro
and in ViVo evaluation.
ACKNOWLEDGMENT
This work was supported by The Methodist Hospital Research
Institute, the M.D. Anderson Foundation, and the Vivian L.
Smith Foundation. GSJ would like to acknowledge support from

278 Bioconjugate Chem., Vol. 21, No. 2, 2010
Li et al.
S. M., Yan, H., Riggins, G. J., Bigner, D. D., Karchin, R.,
Papadopoulos, N., Parmigiani, G., Vogelstein, B., Velculescu,
V. E., and Kinzler, K. W. (2008) An integrated genomic analysis
of human glioblastoma multiforme. Science 321, 1807–12.
(18) Kanu, O. O., Mehta, A., Di, C., Lin, N., Bortoff, K., Bigner,
D. D., Yan, H., and Adamson, D. C. (2009) Glioblastoma
multiforme: a review of therapeutic targets. Expert Opin. Ther.
Targets 13, 701–18.
(19) Rao, J., Lahiri, J., Isaacs, L., Weis, R. M., and Whitesides, G. M.
(1998) A trivalent system from vancomycin.D-ala-D-Ala with higher
affinity than avidin.biotin. Science 280, 708–11.
(20) Mulder, A., Huskens, J., and Reinhoudt, D. N. (2004)
Multivalency in supramolecular chemistry and nanofabrication.
Org. Biomol. Chem. 2, 3409–24.
BC900313D