Magnetic glyco-nanoparticles: A tool to detect, differentiate, and unlock the glyco-codes of cancer via magnetic resonance imaging

Article abstract of DOI:10.1021/ja100455c

Within cancer, there is a large wealth of diversity, complexity, and information that nature has engineered rendering it challenging to identify reliable detection methods. Therefore, the development of simple and effective techniques to delineate the fin

Full text of DOI:10.1021/ja100455c

Published on Web 03/04/2010
Magnetic Glyco-Nanoparticles: A Tool To Detect, Differentiate,
and Unlock the Glyco-Codes of Cancer via Magnetic
Resonance Imaging
Kheireddine El-Boubbou,† David C. Zhu,‡ Chrysoula Vasileiou,† Babak Borhan,†
Davide Prosperi,§ Wei Li,⊥ and Xuefei Huang*,†
Departments of Chemistry, Radiology, and Psychology, Michigan State UniVersity, East Lansing,
Michigan 48824, Department of Biotechnology and Bioscience, UniVersity of Milano-Bicocca,
Piazza della Scienza 2, 20126 Milan, Italy, and College of Pharmacy, UniVersity of Tennessee
Health Science Center, 847 Monroe AVenue, Memphis, Tennessee 38163
Received January 18, 2010; E-mail: xuefei@chemistry.msu.edu
Abstract: Within cancer, there is a large wealth of diversity, complexity, and information that nature has
engineered rendering it challenging to identify reliable detection methods. Therefore, the development of
simple and effective techniques to delineate the fine characteristics of cancer cells can have great potential
impacts on cancer diagnosis and treatment. Herein, we report a magnetic glyco-nanoparticle (MGNP) based
nanosensor system bearing carbohydrates as the ligands, not only to detect and differentiate cancer cells
but also to quantitatively profile their carbohydrate binding abilities by magnetic resonance imaging (MRI).
Using an array of MGNPs, a range of cells including closely related isogenic tumor cells, cells with different
metastatic potential and malignant vs normal cells can be readily distinguished based on their respective
“MRI signatures”. Furthermore, the information obtained from such studies helped guide the establishment
of strongly binding MGNPs as antiadhesive agents against tumors. As the interactions between glyco-
conjugates and endogenous lectins present on cancer cell surface are crucial for cancer development and
metastasis, the ability to characterize and unlock the glyco-code of individual cell lines can facilitate both
the understanding of the roles of carbohydrates as well as the expansion of diagnostic and therapeutic
tools for cancer.
tendencies to mutate,7,8 their antigenic variants may escape the
detection leading to false negative results. An appealing
Introduction
Cancer is a complex group of diseases that affect the lives
of millions of people worldwide. Each malignant cell type has
molecular signatures that distinguish it from its healthy coun-
terpart. The availability of simple and fast methods to identify
the unique cellular characteristics can greatly benefit cancer
treatment and improve the clinical outcomes for patients.1-3
Currently, the majority of methods for cancer detection target
biomarkers such as mutated DNA/RNA and overexpressed
antigens.1,3–5 This requires extensive prior knowledge on the
presence of the specific markers,4,6 which can be very time-
consuming to acquire. Furthermore, as tumor cells have high
alternative is to take advantage of cell surface receptor mediated
recognition events. Receptor binding is often critical to cell
functions and usually cannot be abolished without affecting cell
viability. In addition, cell surface receptors would be easier to
target without requiring the probes to cross the cellular
membrane as compared to intracellular markers.
An attractive target for receptor-mediated interaction is
carbohydrates and, in particular glycoconjugates, which play
important roles in cancer development and metastasis.9-12
Carbohydrates are uniquely suited for encoding biological
information because of their rich structural variations.13-15
Aberrant glycosylations on tumor cell surfaces have been
† Department of Chemistry, Michigan State University.
‡ Departments of Radiology and Psychology, Michigan State University.
§ University of Milano-Bicocca.
(7) Hynes, N. E.; MacDonald, G. Curr. Opin. Cell Biol. 2009, 21, 177–
184.
⊥ University of Tennessee Health Science Center.
(1) Lee, H.; Yoon, T.-J.; Figueiredo, J.-L.; Swirski, F. K.; Weissleder, R.
Proc. Natl. Acad. Sci. U.S.A. 2009, 106, 12459–12464.
(2) Bajaj, A.; Miranda, O. R.; Kim, I.-B.; Phillips, R. L.; Jerry, D. J.;
Bunz, U. H. F.; Rotello, V. M. Proc. Natl. Acad. Sci. U.S.A. 2009,
106, 10912–10916, and references therein.
(8) Klein, C. A.; Blankenstein, T. J. F.; Schmidt-Kittler, O.; Petronio,
M.; Polzer, B.; Stoecklein, N. H.; Riethmu¨ller, G. Lancet 2002, 360,
683–689.
(9) Freire, T.; Bay, S.; Vichier-Guerre, S.; Lo-Man, R.; Leclerc, C. Mini-
ReV. Med. Chem. 2006, 6, 1357–1373.
(3) Nagrath, S.; Sequist, L. V.; Maheswaran, S.; Bell, D. W.; Irimia, D.;
Ulkus, L.; Smith, M. R.; Kwak, E. L.; Digumarthy, S.; Muzikansky,
A.; Ryan, P.; Balis, U. J.; Tompkins, R. G.; Haber, D. A.; Toner, M.
Nature 2007, 450, 1235–1239.
(10) Hakomori, S. Proc. Natl. Acad. Sci. U.S.A. 2002, 99, 10231–10233.
(11) Danishefsky, S. J.; Allen, J. R. Angew. Chem., Int. Ed. 2000, 39, 836–
863, and references therein.
(12) Hakomori, S.; Zhang, Y. Chem. Biol. 1997, 4, 97–104.
(13) Gabius, H.-J. Biochem. Soc. Trans. 2008, 36, 1491–1496.
(14) Gabius, H.-J.; Siebert, H.-C.; Andre´, S.; Jime´nez-Barbero, J.; Ru¨diger,
H. ChemBioChem 2004, 5, 740–764.
(4) Pantel, K.; Brakenhoff, R. H.; Brandt, B. Nat. ReV. Cancer 2008, 8,
329–340.
(5) Ludwig, J. A.; Weinstein, J. N. Nat. ReV. Cancer 2005, 5, 845–856.
(6) Ferrari, M. Nat. ReV. Cancer 2005, 5, 161–171.
(15) Varki, A. Glycobiology 1993, 3, 97–130.
9
4490 J. AM. CHEM. SOC. 2010, 132, 4490–4499
10.1021/ja100455c  2010 American Chemical Society

Magnetic Glyco-Nanoparticles To Detect Tumor Cells
A R T I C L E S
extensively probed by antibodies and plant-derived lectins.12,16
This led to the identification of characteristic tumor-associated
carbohydrate molecules,11,12 which has greatly facilitated the
development of carbohydrate-based anticancer vaccine studies.9,17
In comparison, the understanding of carbohydrate-binding
properties of tumors is not as advanced. Cancer cells can interact
with the extracellular matrix in their microenvironment through
endogenous receptors binding with carbohydrates.18-20 These
interactions vary, depending on the physiological state of the
cells, as supported by the ground-breaking histological studies
of tumor tissues.21-23 Therefore, the ability to characterize and
distinguish carbohydrate binding profiles of a variety of cells
can expedite both the mechanistic understanding of their roles
in disease development and the expansion of diagnostic and
therapeutic tools.24-26 As the distinctions among cancer cell
subtypes and malignant vs normal cells can often be subtle, a
suitable tool is needed to quantitatiVely analyze the fine
characteristics in carbohydrate binding of various cell types.
MGNPs can allow cell detection via magnetic resonance
imaging (MRI) without the need to prelabel the cells.31–33
One challenge, however, in using MGNPs and carbohydrates
for molecular recognition is that multiple cell types may bind
with the same carbohydrate structure albeit in different affinities.
To address this issue, we envision that by pooling the responses
from an array of MGNPs, the various cell types may be
differentiated through pattern recognition.2,34-36 Furthermore,
the information obtained on the physiologically relevant
carbohydrate-receptor interaction can not only enhance our
understanding of the roles carbohydrate play in cancer but also
guide the development of potential therapeutics such as agents
against cancer adhesion. Although glyco-nanoparticles have been
previously employed for elegant studies of carbohydrate-
mediated biological recognitions,37-48 MGNPs have not been
utilized to detect and systematically profile mammalian cells.
Experimental Section
Cells and Culture Conditions. Unless otherwise indicated, all
starting materials, reagents and solvents were obtained from
commercial suppliers (Sigma-Aldrich or Fisher Scientific) and used
as supplied without further purification. All fluorescein-labeled
lectins were purchased from Aldrich. All cell lines were purchased
from the American Type Culture Collection (ATCC) [cell line
designation (catalogue no.), type] unless otherwise noted: 184B5
(CRL-8799), normal breast cell; A498 (HTB-44), kidney cancer;
A549 (CCL-185), lung cancer; HT29 (HTB-38), colon cancer;
SKOV-3 (HTB-77), ovarian cancer; B16-F10 (CRL-6475), meta-
static mouse melanoma; B16-F1 (CRL-6323), less metastatic mouse
melanoma. The MCF-7/Adr-res (adriamycin-resistant breast cancer)
cell line was obtained from the National Cancer Institute. Two
murine mammary carcinoma cell lines (TA3-HA, TA3-ST) were
kind gifts from Dr. John Hilkens, Netherlands Cancer Institute. All
In the past decade, nanotechnology has begun to play
increasingly important roles in cancer research.6 Using antibody-
immobilized nanoparticles, various types of cancer cells were
detected both in Vitro1 and in ViVo.27 Recently, instead of relying
on the specific antibodies, structurally related cationic gold
nanoparticles bearing fluorescent polymers on the surface were
prepared.2 The differential electrostatic and hydrophobic interac-
tions between the gold nanoparticles and cells were reflected
in changes of fluorescence intensity upon cell binding, which
allowed the differentiation of tumor from normal cells as well
as closely related tumor cells. Herein, we explore the possibility
of using a magnetic glyco-nanoparticle (MGNP)-based system
to detect and profile various cell types on the basis of the more
physiologically related carbohydrate-receptor interactions.
MGNPs provide an appealing platform for biological detection.
The spherical nanoparticles have large surface areas, which
allow the attachment of multiple carbohydrates leading to
enhanced avidity with carbohydrate receptors through multiva-
lent binding.13,28 Unlike the toxic heavy metal-containing
nanoparticles such as quantum dots,29,30 the magnetite nano-
particles have been approved for clinical uses with minimum
cytotoxicity.31 Furthermore, the superparamagnetic nature of this
(29) Chang, E.; Thekkek, N.; Yu, W. W.; Colvin, V. L.; Drezek, R. Small
2006, 2, 1412–1417.
(30) Derfus, A. M.; Chan, W. C. W.; Bhatia, S. N. Nano Lett. 2004, 4,
11–18.
(31) Gupta, A. K.; Gupta, M. Biomaterials 2005, 26, 3995–4021.
(32) Wang, Y.-X. J.; Hussain, S. M.; Krestin, G. P. Eur. Radiol. 2001, 11,
2319–2331.
(33) Weissleder, R.; Elizondo, G.; Wittenberg, J.; Rabito, C. A.; Bengele,
H. H.; Josephson, L. Radiology 1990, 175, 489–493.
(34) Collins, B. E.; Wright, A. T.; Anslyn, E. V. Top. Curr. Chem. 2007,
277, 181–218.
(35) Wright, A. T.; Anslyn, E. V. Chem. Soc. ReV. 2006, 35, 14–28.
(36) Jurs, P. C.; Bakken, G. A.; McClelland, H. E. Chem. ReV. 2000, 100,
2649–2678.
(16) Dai, Z.; Zhou, J.; Qiu, S.-J.; Liu, Y.-K.; Fan, J. Electrophoresis 2009,
30, 2957–2966, and references therein.
(17) Buskas, T.; Thompson, P.; Boons, G.-J. Chem. Commun. 2009, 5335–
5349.
(37) Barrientos, A. G.; de la Fuente, J. M.; Jime´nez, M.; Sol´ıs, D.; Can˜ada,
F. J.; Mart´ın-Lomas, M.; Penade´s, S. Carbohydr. Res. 2009, 344,
1474–1478.
(18) Barthel, S. R.; Gavino, J. D.; Descheny, L.; Dimitroff, C. J. Exp. Opin.
Ther. Targets 2007, 11, 1473–1491.
(38) Liang, C.-H.; Wang, C.-C.; Lin, Y.-C.; Chen, C.-H.; Wong, C.-H.;
Wu, C.-Y. Anal. Chem. 2009, 81, 7750–7756.
(19) Kannagi, R.; Izawa, M.; Koike, T.; Miyazaki, K.; Kimura, N. Cancer
Sci. 2004, 65, 377–384.
(39) Marradi, M.; Alca´ntara, D.; de la Fuente, J. M.; Garc´ıa-Mart´ın, M. L.;
Cerda´n, S.; Penade´s, S. Chem. Commun. 2009, 3922–3924.
(40) van Kasteren, S. I.; Campbell, S. J.; Serres, S.; Anthony, D. C.; Sibson,
N. R.; Davis, B. G. Proc. Natl. Acad. Sci. U.S.A. 2009, 106, 18–23.
(41) Kikkeri, R.; Lepenies, B.; Adibekian, A.; Laurino, P.; Seeberger, P. H.
J. Am. Chem. Soc. 2009, 131, 2110–2112.
(20) Lahm, H.; Andre, S.; Hoeflich, A.; Kaltner, H.; Siebert, H.-C.; Sordat,
B.; von der Lieth, C.-W.; Wolf, E.; Gabius, H.-J. Glycoconjugate J.
2004, 20, 227–238.
(21) Gabius, H.-J. Anat. Histol. Embryol. 2001, 30, 3–31, and references
therein.
(22) Kayser, K.; Bovin, N. V.; Zemlyanukhina, T. V.; Donaldo-Jacinto,
S.; Koopmann, J.; Gabius, H.-J. Glycoconjugate J. 1994, 11, 339–
344.
(42) Thygesen, M. B.; Sauer, J.; Jensen, K. J. Chem.sEur. J. 2009, 15,
1649–1660.
(43) Mukhopadhyay, B.; Martins, M. B.; Karamanska, R.; Russell, D. A.;
Field, R. A. Tetrahedron Lett. 2009, 50, 886–889.
(44) Sundgren, A.; Barchi, J. J. Carbohydr. Res. 2008, 343, 1594–1604.
(45) El-Boubbou, K.; Gruden, C.; Huang, X. J. Am. Chem. Soc. 2007, 129,
13392–13393.
(23) Kayser, K.; Heil, M.; Gabius, H.-J. Pathol., Res. Pract. 1989, 184,
621–629.
(24) Kim, E. Y. L.; Gronewold, C.; Chatterjee, A.; Von der Lieth, C.-W.;
Kliem, C.; Schmauser, B.; Wiessler, M.; Frei, E. ChemBioChem 2005,
6, 422–431 and references cited therein.
(46) de Souza, A. C.; Halkes, K. M.; Meeldijk, J. D.; Verkleij, A. J.;
Vliegenthart, J. F. G.; Kamerling, J. P. ChemBioChem 2005, 6, 828–
831.
(25) Lerchen, H.-G.; Baumgarten, J.; Piel, N.; Kolb-Bachofen, V. Angew.
Chem., Int. Ed. 1999, 38, 3680–3683.
(26) Rye, P. D.; Bovin, N. V. Glycobiology 1997, 7, 179–182.
(27) Gao, X.; Cui, Y.; Levenson, R. M.; Chung, L. W. K.; Nie, S. Nat.
Biotechnol. 2004, 22, 969–979.
(47) Hone, D. C.; Haines, A. H.; Russell, D. A. Langmuir 2003, 19, 7141–
7144.
(48) Lin, C.-C.; Yeh, Y.-C.; Yang, C.-Y.; Chen, G.-F.; Chen, Y.-C.; Wu,
Y.-C.; Chen, C.-C. Chem. Commun. 2003, 2920–2921.
(28) Lee, Y. C.; Lee, R. T. Acc. Chem. Res. 1995, 28, 321–327.
9
J. AM. CHEM. SOC. VOL. 132, NO. 12, 2010 4491

A R T I C L E S
El-Boubbou et al.
Scheme 1. Synthesis of MGNPs
cell lines were grown as monolayer cultures on tissue culture dishes
in phenol red free DMEM (Invitrogen, Carlsbad, CA) supplemented
with 10% fetal bovine serum (FBS) (Sigma), penicillin G (Sigma,
61.4 µg/mL), streptomycin (Sigma, 100 µg/mL) and L-glutamine
(Sigma, 292 µg/mL) at 37 °C in an atmosphere of 5% CO2 and
95% air. All cells were grown to log phase, trypsinized with
trypsin-EDTA solution (0.25% trypsin, 1 mM EDTA) to detach
the cells, washed twice by centrifugation to remove any residual
trypsin, and resuspended in the appropriate media.
cells for each MGNP was set as 100%. The ∆T2% for each cell
line was calculated as the percentage of change relative to the largest
T2 changes for the respective MGNP. LDA was performed using
the statistical computing and graphics software R.
Prussian Blue Staining. Different cancer cell lines were seeded
onto 24-well plate. After incubation for 12 h at 37 °C, nanoparticles
were added to the plate in a final concentration of 20 µg/mL per
well. After 12 h, the supernatant was removed, and cells were
washed three times with PBS, treated with 10% formalin solution
(0.5 mL) for 5 min to fix the cells, and then washed with PBS.
Prussian blue staining was then performed. To each well was added
a 1:1 mixture of 4% potassium ferrocyanide(II) trihydrate and 2%
HCl solution (0.5 mL), and cells were incubated for 30 min at 37 °C
in the dark, counterstained with nuclear fast red for 3 min, and
then washed three times with PBS. The Prussian blue staining
images were assessed by an inverted light microscope (Figures 6
and S9 (Supporting Information)).
Antiadhesive Assay. Cancer cells were detached using 2 mM
EDTA in PBS, washed with PBS, counted to final concentration
of ∼5 × 104 cells/mL and incubated with Gal-MGNP (20 µg/mL)
for 3 h at 37 °C in DMEM containing 10% FBS. The cells were
then immediately seeded in a 24-well plate and incubated at 37 °C.
After 10, 20, 30, and 45 min of incubation, the medium and the
floating cells were carefully removed by aspiration, and the attached
layers were washed twice with PBS. The firmly attached cells were
then counted under an inverted light microscope. Cell adhesion
curves were generated after counting triplicate wells (20 different
homogeneous fields per each well). The same experiment was also
performed at a different cell density of ∼5 × 105 cells/mL (Figure
S11, Supporting Information).
HR-MAS 1H NMR Analyses. HR-MAS experiments were
carried out on a Bruker BioSpin FT-NMR Avance 500 equipped
with an 11.7 T superconducting ultrashield magnet available at
C.IGA (Centro Interdipartimentale Grandi Apparecchiature) of the
University of Milan. The HR-MAS probe with internal lock is
capable of performing either direct or indirect (inverse) detection
experiments. MAS experiments were performed at spinning rates
of up to 15 kHz (15 kHz maximum MAS rotation available) using
a 50 µL zirconia rotor. All the samples were diluted at different
concentrations with deuterated solvents to find out the concentration
1
limit to the NMR signal broadening. HR-MAS H NMR spectra
were obtained using 200-400 scans for each experiment. The
sample temperature was dependent on the rotation speed.
MRI and Relaxivity Measurement. Magnetic resonance imag-
ing studies were carried on a 3T Signa HDx MR scanner (GE
Healthcare, Waukesha, WI). Thirty test tube samples were scanned
simultaneously in a standard quadrature birdcage head coil. For
T2 measurements, a multiecho fast spin-echo sequence (time of
repetition ) 500 ms, receiver bandwidth ) ( 31.25 kHz, field of
view ) 20 cm, slice thickness ) 3 mm, gap ) 3 mm, number
of excitation ) 1, and matrix size ) 256 × 128) was used to
simultaneously collect a series of data points at seven different echo
times (TE) 15-60 ms with an increment of 7.5 ms at two slice
locations. For each sample, the region of interest (ROI) (circles of
4.5 mm radius) was drawn at the center of each test tube at both
slice locations. The T2 was calculated on the basis of the semilog
linear regression of the mean signal intensity values at the ROI
and the corresponding TEs. Specifically, 1/T2 ) -[(ln Sn - ln Sm)/
(TEn - TEm)], where Sn and Sm are voxel signal intensity values at
TE values of TEn and TEm.
Results and Discussion
Synthesis and Characterization of MGNPs. The synthesis of
MGNPs commenced from the tetraethoxysilane coated magne-
tite nanoparticle (NP 1),45,49 on which amino groups were
introduced via silanization with 3-aminopropyl triethoxysilane
(APTES) (Scheme 1a). Carboxylic acid derivatives of four types
of naturally occurring monosaccharides, namely, mannose
(Man), galactose (Gal), fucose (Fuc), and sialic acid (Sia), were
then immobilized onto the amine-functionalized nanoparticles
through amide coupling reactions leading to MGNP 2-5,
respectively (Scheme 1a).
Detection of Cancer Cells using MGNPs. Cell suspensions (105
or 106 cells/mL) were prepared in phenol red free DMEM media
supplemented with 0.2% bovine serum albumin. Aliquots of these
cultures were placed in sterile tubes, and MGNPs 2-6 (final
concentration 20 µg/mL) were added. NP 1 without any carbohy-
drates was used as the control. The tubes were incubated with gentle
mixing at 37 °C. The T2 values of MGNP/cell suspensions and
MGNP in the absence of cells were then recorded via MRI. MRI
experiments were performed eight times for each cell line at both
cell concentrations. The largest T2 value change upon binding with
In an alternative method, we synthesized GlcNAc-MGNP 6
via click chemistry. To this end, we explored the possibility of
using a native unprotected sugar. The free reducing sugar
(49) Srinivasan, B.; Huang, X. Chirality 2008, 20, 265–277.
9
4492 J. AM. CHEM. SOC. VOL. 132, NO. 12, 2010

Magnetic Glyco-Nanoparticles To Detect Tumor Cells
A R T I C L E S
Scheme 2a
lectins, i.e., ConcanaValin A (Con A, a Man selective lectin),56
Wheat Germ Agglutinin (WGA, a GlcNAc and Sia selective
lectin),57,58 a Bandeiraea simplicifolia isolectin (BS-I, a Gal
selective lectin)59 and Lotus Tetragonolobus purpureas Ag-
glutinin (TPA, a Fuc-selective lectin).60 Upon incubation of a
fluorescently labeled lectin with a MGNP, if the MGNP can
bind with the lectin, subsequent application of an external
magnetic field to the sample would remove the fluorescent lectin
from the solution, leading to a reduction of fluorescent intensity
of the supernatant. (For complete results on all five MGNPs
binding with the four lectins, see Figures S4 and S5 (Supporting
Information).) Con A is a well-characterized Man selective lectin
with weak Gal-binding affinities.56 Incubation of Man MGNP
2 (1 mg/mL) with fluorescein isocyanate (FITC)-labeled Con
A (100 µg/mL) followed by magnetic separation led to a 89%
reduction in fluorescent intensity of the solution, while the same
amount of the weakly bound Gal-MGNP 3 was able to remove
only 8% of the Con A (Figure 2a). This is consistent with the
known carbohydrate binding preferences of Con A.56 The
addition of a solution of free mannose (18 mg/mL, 100 mM)
to Man-MGNP 2/Con A mixture did not increase the intensity
of residual emission of the supernatant after magnetic separation
and high concentration of mannose (180 mg/mL, 1 M) was
required to partially disrupt the Con A/Man-MGNP 2 complex.
These results reveal that multivalent display of carbohydrate
ligands on MGNP resulted in strong lectin binding.
a Reagents and conditions: a) Benzoyl chloride, pyridine, CH2Cl2, 0 °C
to rt, 2 h; b) 3-bromopropyne, K2CO3, acetone, reflux for 8 h; c) 6% MeOH
in HCl, reflux for 2 h; d) N-acetyl glucosamine, 0.1 M acetate buffer (pH
) 6.5)/DMF (3:1), 50 °C, 24 h.
N-acetyl glucosamine (GlcNAc) was chemoselectively ligated
with a methoxy amine linker producing GlcNAc derivative 7
containing a terminal alkyne at the reducing end (Scheme 2).50,51
NP 1 was then modified with the azido siloxane derivative 8
(see Scheme S4 for synthetic details (Supporting Information)),
which was subsequently coupled with alkynyl-GlcNAc 7
through the copper-catalyzed Huisgen click reaction52-54 to
yield GlcNAc-MGNP 6 (Scheme 1b). It is advantageous to use
a native unprotected sugar as this opens up an avenue for future
incorporation of natural polysaccharides without extensive
synthetic manipulations.
All NPs assembled were characterized by a variety of
techniques, including transmission electron microscopy (TEM),
nuclear magnetic resonance (NMR), thermogravimetric analysis
(TGA), infrared spectroscopy, and X-ray diffraction (Figures
S1-S3, Supporting Information). TEM images of MGNPs
indicated that the diameters of NPs were around 6 nm (Figure
1a) and TGA showed that about 8% of the dry weight of
MGNPs was due to carbohydrates (Figure S1, Supporting
Information). NMR is a critical tool for providing detailed ligand
structural features. Unfortunately, there is a serious limitation
directly applying 1H NMR to MGNPs due to the field inhomo-
geneity caused by the inherent superparamagnetism of the
particles. Drastic line broadening was observed in spectra
recorded with a conventional NMR probe, which led to
undistinguishable 1H NMR spectra (Figure 1b). Recently, high
resolution-magic angle spinning (HR-MAS) NMR was found
to be a superior tool to overcome this problem.52,55 Indeed, HR-
In contrast to Con A, the Gal-selective FITC-BS-I59 strongly
bound with Gal-MGNP 3, producing 91% reduction of the
solution emission intensity, which was competitively inhibited
by a concentrated solution of free galactose (Figure 2b). At the
same time, as BS-I has weak affinities with mannose,59
incubation of Man-MGNP 2 with FITC-BS-I only decreased
emission intensity a little. The same phenomena were observed
with WGA and TPA (Figure S4, Supporting Information). WGA
bound tightly with GlcNAc-MGNP 6 and Sia-MGNP 5 but not
with Gal-MGNP 3 and TPA interacted strongly with Fuc-MGNP
4 as predicted on the basis of their known binding spec-
ificities.57,58,60,61 In addition, NP 1 devoid of carbohydrates on
the surface could not remove any proteins from the solution,
suggesting that the nonspecific absorption on NP surface is
minimal.
To further validate the specificity, we investigated the
interactions of MGNP with a well-characterized Escherichia
coli system. The E. coli strain ORN178 contains the mannose
binding protein FimH in its fimbriae, while the ORN208 strain
has its FimH mutated, resulting in much reduced mannose
affinity.62 Upon incubation of the ORN178 strain with Man-
MGNP 2 and subsequent magnetic separation, close to 85% of
the bacteria was removed from the media (Figure S6, Supporting
Information).45 In comparison, the mutant strain ORN208 bound
with Man-MGNP 2 weakly, only losing 8% of the cells.
Meanwhile, the Gal-MGNP 3 was ineffective in removing either
strain. These observations coupled with the lectin experiments
unequivocally demonstrated that the MGNPs are not promiscu-
1
MAS H NMR spectra of MGNPs gave solution- like spectra
with a completely resolved splitting pattern, including the correct
signal multiplicities and accurate integrations (Figure 1c). For
instance, the anomeric proton of the galactoside ligand on Gal-
MGNP 3 was well resolved as a doublet (J ) 8.0 Hz) at 4.3
ppm, indicating the configuration of the carbohydrate was
unaffected during immobilization. The fact that only one set of
peaks from carbohydrates was observed suggested that the
carbohydrate coating was homogeneous on the particle surface.
Validation of MGNP-Binding Specificities. Although glyco-
nanoparticles have been previously utilized to probe carbo-
hydrate-receptor interactions,37–48 it is important to validate
that the carbohydrates immobilized on MGNPs retain their
biological recognition specificity. This was probed using four
(50) Bohorov, O.; Andersson-Sand, H.; Hoffmann, J.; Blixt, O. Glycobi-
ology 2006, 16, 21C–27C.
(56) Shimura, K.; Kasai, K. Anal. Biochem. 1995, 227, 186–194.
(57) Nagata, Y.; Burger, M. M. J. Biol. Chem. 1974, 249, 3116–3122.
(58) Vornholt, W.; Hartmann, M.; Keusgen, M. Biosen. Bioelectron. 2007,
22, 2983–2988.
(51) Peri, F.; Jime´nez-Barbero, J.; Garc´ıa-Aparicio, V.; Tvarosˇka, I.;
Nicotra, F. Chem.sEur. J. 2004, 10, 1433–1444.
(52) Polito, L.; Monti, D.; Caneva, E.; Delnevo, E.; Russo, G.; Prosperi,
D. Chem. Commun. 2008, 621–623.
(59) Hayes, C. E.; Goldstein, I. J. J. Biol. Chem. 1974, 249, 1904–1914.
(60) Liener, I. E., Sharon, N., Goldstein, I. J., Eds. The Lectins: Properties,
Functions, and Applications in Biology and Medicine; Academic Press:
Orlando, FL, 1986.
(53) Lin, P.-C.; Ueng, S.-H.; Yu, S.-C.; Jan, M.-D.; Adak, A. K.; Yu, C.-
C.; Lin, C.-C. Org. Lett. 2007, 9, 2131–2134.
(54) Wang, Q.; Chan, T. R.; Hilgraf, R.; Fokin, V. V.; Sharpless, K. B.;
Finn, M. G. J. Am. Chem. Soc. 2003, 125, 3192–3193.
(55) Polito, L.; Colombo, M.; Monti, D.; Melato, S.; Caneva, E.; Prosperi,
D. J. Am. Chem. Soc. 2008, 130, 12712–24.
(61) Wright, C. S. J. Mol. Biol. 1984, 178, 91–104.
(62) Harris, S. L.; Spears, P. A.; Havell, E. A.; Hamrick, T. S.; Horton,
J. R.; Orndorff, P. E. J. Bacteriol. 2001, 183, 4099–4102.
9
J. AM. CHEM. SOC. VOL. 132, NO. 12, 2010 4493

A R T I C L E S
El-Boubbou et al.
1
Figure 1. (a) TEM of Gal-MGNP 3 (the scale bar is 5 nm); (b) H NMR spectrum of Gal-MGNP 3 acquired in solution with a conventional 5 mm probe
1
(QNP); and (c) HR-MAS H NMR of the same sample.
Figure 2. Fluorescent emission intensities of supernatants of (a) FITC-Con A (100 µg/mL) and (b) FITC-BS-I (100 µg/mL) upon incubation with various
MGNPs (1 mg/mL) followed by magnet-mediated separation (λexcitation ) 494 nm). Man-MGNP 2 bound strongly with the Man-selective Con A but not with
Gal-selective BS-I, while Gal-MGNP 3 bound tightly with BS-I but not with Con A. The binding between lectins and the specfic MGNP could be competitively
inhibited with a concentrated solution of the same carbohydrate.
ous in binding and the carbohydrates immobilized on MGNPs
maintain the same biological recognition preference as the free
forms in solution.
to the small sizes of MGNPs, multiple MGNPs can bind with
the receptor, assembling into clusters (Figure 3). With their
increased sizes, the aggregates create larger local magnetic field
gradients and thus become more efficient at dephasing the spins
of surrounding water protons as the motional averaging condi-
tion is satisfied at the nanoparticle size regime.66,67 This leads
to lowering of T2 relaxation times as described by the magnetic
relaxation switching theory.1,63,68
Our detection assay was first tested using Con A. As one
Con A contains four mannose binding sites,69 it can cross-link
multiple mannose containing Man-MGNP 2, leading to NP
aggregation, which should result in the reduction of T2
relaxation time (Figure 3a). Due to the superparamagnetic nature
of the magnetic nanoparticles, only 20 µg/mL of MGNP was
needed for detection. When Man-MGNP 2 was mixed with
increasing concentrations of Con A, the binding equilibrium
was shifted more toward the aggregates. This led to a sequential
decrease of the brightness of the corresponding T2-weighted
Monitoring MGNP Binding by MRI. After establishing the
specificity of MGNPs using fluorescence experiments, we
moved on to examine the utility of MRI to monitor the
interaction of MGNPs with their biological targets. MRI
measures the relaxation time of water protons in a magnetic
field, which is commonly used to noninvasively visualize the
internal structures and functions in ViVo. Using MRI to monitor
MGNP allows multiple samples to be measured simultaneously
within a single scan,63 thus enabling rapid response time and
shortening multianalyte data acquisition. Furthermore, when the
MRI methodology for MGNP monitoring is established, it can
be translated to in ViVo applications without the need to develop
a new detection technology.
Magnetic NPs can serve as MRI contrast agents, where they
decrease the transverse relaxation time (T2), producing a
negative contrast from the environment by virtue of signal
reduction.32,64,65 In the presence of a cross-linking receptor, due
(66) Rocha, A.; Gossuinb, Y.; Mullera, R. N.; Gillis, P. J. Magn. Magn.
Mater. 2005, 293, 532–539.
(63) Lee, H.; Sun, E.; Ham, D.; Weissleder, R. Nat. Med. 2008, 14, 869–
874.
(67) Yung, K.-T. Magn. Reson. Imaging 2003, 21, 451–463.
(68) Koh, I.; Hong, R.; Weissleder, R.; Josephson, L. Angew. Chem., Int.
Ed. 2008, 47, 4119–4121.
(64) Thorek, D. L. J.; Chen, A. K.; Czupryna, U.; Tsourkas, A. Ann.
Biomed. Eng. 2005, 34, 23–38.
(69) Derewenda, Z.; Yariv, J.; HelliweII, J. R.; Kalb, A. J.; Dodson, E. J.;
Papiz, M. Z.; Wan, T.; Campbell, J. EMBO J. 1989, 8, 2189–2193.
(65) Weissleder, R.; Papisov, M. ReV. Magn. Reson. Med 1992, 4, 1–20.
9
4494 J. AM. CHEM. SOC. VOL. 132, NO. 12, 2010

Magnetic Glyco-Nanoparticles To Detect Tumor Cells
A R T I C L E S
Figure 3. (a) Incubation of Man-MGNP 2 (shown as pink balls) with Con A, a tetrameric mannose selective lectin (shown as blue rectangles) resulted in
the formation of aggregates, leading to shorter T2 relaxation time and consequently a darkened MRI image. (b) T2-weighted MRI images of Man-MGNP
2 (20 µg/mL) and NP 1 (20 µg/mL) upon incubation with increasing concentrations of Con A. (c) T2 changed linearly upon incubation of Man-MGNP 2
with increasing concentrations of Con A, while incubation with NP 1 devoid of carbohydrates did not lead to any change. (d) T2 changed linearly upon
incubation of GlcNAc-MGNP 6 with increasing concentrations of the GlcNAc-selective lectin WGA, while incubation with the nonbinding Gal-MGNP 3
did not cause any T2 change.
MR images (Figure 3b). Quantification of the images showed
an excellent linear correlation between Con A concentration and
T2 relaxation time, with as little as 0.1 µg/mL (1 nM) of Con
A detected (Figure 3c). In contrast, the control NP 1 devoid of
mannose did not cause any change in T2, signifying that the
contrast change was due to the specific binding between Man-
MGNP 2 and Con A. In addition, incubation of the GlcNAc
selective WGA with GlcNAc-MGNP 6 led to a linear decrease
of T2 relaxation time with increasing concentrations of WGA,
while the mixture of WGA with the nonbinding Gal-MGNP 3
did not produce any T2 changes (Figure 3d). These results
corroborated the fluorescence studies and confirmed MGNP
binding specificities (Figures 2, S4, S6, and S7 (Supporting
Information)), which gave us great confidence to apply this
technology to cancer cell study.
Mammalian Cells Selectively Bound with MGNPs As De-
tected by MRI. Building on the success of lectin and E. coli
detection, we evaluated the utility of MGNPs in monitoring
mammalian cell interactions and cancer cell detection. The use
of carbohydrates as the recognition elements can provide
functional information on cell surface active carbohydrate
receptors. This is complementary to the usage of antibodies, as
the latter monitor the presence of particular antigenic structures,
which can be absent in some cancer cell mutants. In addition,
an antibody is commonly limited to binding a specific target,
while carbohydrate ligands can be used to monitor a range of
cells, thus reducing the number of reagents required for study.
of cells at two concentrations (105 and 106 cells/mL) was
incubated with MGNP 2-6 or the control NP 1, and the T2
relaxation times of all the samples were recorded. Thirty samples
were measured at the same time with our MRI setup. While no
significant T2 changes were observed with any cells upon mixing
with the NP 1, the 10 cell lines produced a large variation in
T2 reductions upon MGNP incubation with the T2 changes
normalized against the largest ∆T2 within each MGNP category
(Figure 4).
The decrease of the absolute values of T2 upon MGNP
incubation can be explained due to particle agglomeration upon
cell binding. At a higher cell concentration (106 cells/mL, Figure
4b), more MGNPs were bound leading to larger ∆T2 compared
with lower cell concentration (105 cells/mL) for each cell line
(Figure 4a). Furthermore, when the cells were preincubated with
a solution of a free monosaccharide (100 mM) and then treated
with the MGNP bearing the same carbohydrate, T2 changes were
less due to the competitive binding of the free monosaccharide
with cells. These observations were consistent with the notion
that T2 reduction was induced by the specific MGNP/cell
interactions.
Biological Implications of Carbohydrate-Receptors on Can-
cer Cells. Based on the MR responses, in most of the cell lines
examined, bindings with Fuc-MGNP 4 and Sia-MGNP 5 were
observed (Figure 4), suggesting these cell lines have active
fucose and sialic acid receptors. B16F1, B16F10, MCF-7/Adr-
res, and SKOV-3 were found to interact with ꢀ-galactoside. This
is of special interest since it confirmed the previously reported
galactoside binding of the B16F1024 presumably through
galectins, a family of galactose specific lectins as well as the
A normal breast cell line 184B5 and nine types of representa-
tive cancer cells were used for our study including human
ovarian adenocarcinoma SKOV-3, colon HT29, kidney A498,
lung A549, and breast cancer MCF-7/Adr-res and the closely
related murine melanoma cell lines B16F10 and B16F1,
mammary adenocarcinoma TA3-ST and TA3-HA. Each type
9
J. AM. CHEM. SOC. VOL. 132, NO. 12, 2010 4495

A R T I C L E S
El-Boubbou et al.
Figure 4. (a) Percentage changes of T2 relaxation time (% ∆T2) obtained upon incubating MGNPs 2-6 or the control NP 1 (20 µg/mL) with 10 cell lines
(105 cells/mL). The ∆T2 was calculated by dividing the T2 differences between MGNP and MGNP/cancer cell by the corresponding highest ∆T2 from each
MGNP category. (b) % ∆T2 obtained upon incubating MGNPs 2-6 or the control NP 1 (20 µg/mL) with 10 cell lines (106 cells/mL). The above data
represent the averages of eight individual measurements with the error bars showing standard deviations.
high expression level of galectins on MCF-7/Adr-res70 cells.
Furthermore, HT29, MCF-7/Adr-res, 184B5, A498, and SK-
OV-3 express functioning GlcNAc receptors as suggested by
their interactions with GlcNAc-MGNP 6. Moreover, MCF-7/
Adr-res, 184B5, SKOV-3, B16F10, B16F1, A549, and A498
bind mannose.71 The sugar-free nanoprobe NP 1 did not bind
to any cells in our study, evidence to the importance of intrinsic
carbohydrate-protein interactions involved and the exclusion
of nonspecific interactions as the cause of binding. The wealth
of new information generated on the sugar-binding prefer-
ences of these cell lines can be very useful for cancer research.
Although some of the receptors responsible for binding are not
determined yet, the magnetic nature of MGNPs can help
facilitate the enrichment and identification of carbohydrate
receptors on these cells in the future through magnet-mediated
separation.
accomplish this, linear discriminant analysis (LDA), a statistical
method for classification of groups of objects, was employed.35
LDA converts the ∆T2 values of each cell line to canonical
linear discriminants (LDs), which are linear combinations of
the original data weighted by coefficients producing the greatest
analyte discrimination. LDA is a powerful technique, which has
been successfully applied to the detection of a variety of targets
including carbohydrates, proteins, and cells.2,35,36,72-74 All T2
changes (6 types of NPs, 10 cell types, 8 repetitions) at each
cell concentration were submitted to LDA, and LDs were
generated. On the basis of the LDA patterns, the 10 cell lines
were easily clustered into 10 respective groups (Figure 5). At
the 105 cells/mL concentration, the first three LDs contain 49.5,
25.3 and 17.5% of the variations respectively, which account
for 92.3% of the total variations. Validation of the LDA was
carried out using a jackknife matrix method,75 where all but
one measurement out of each group was treated as a new
training set. The group memberships of the omitted observations
Establishment of MR Responses as Molecular Signatures
for Full Differentiation of All 10 Cell Lines via LDA Analysis.
With the diverse MR signature in hand, we examined whether
it was possible to differentiate all 10 cell lines. This was a
particularly stringent test due to the large number of cell lines
being analyzed using only five types of MGNPs. In order to
(72) Zhang, T.; Edwards, N. Y.; Bonizzoni, M.; Anslyn, E. V. J. Am. Chem.
Soc. 2009, 131, 11976–11984.
(73) Jagt, R. B. C.; Gomez-Biagi, R. F.; Nitz, M. Angew. Chem., Int. Ed.
2009, 48, 1995–1997.
(74) Miranda, O. R.; You, C.-C.; Phillips, R.; Kim, I.-B.; Ghosh, P. S.;
Bunz, U. H. F.; Rotello, V. M. J. Am. Chem. Soc. 2007, 129, 9856–
9857.
(70) Satelli, A.; Rao, P. S.; Gupta, P. K.; Lockman, P. R.; Srivenugopal,
K. S.; Rao, U. S. Oncol. Rep. 2008, 19, 587–594.
(71) Chandrasekaran, S.; Tanzer, M. L.; Giniger, M. S. J. Biol. Chem. 1994,
269, 3367–73.
(75) Yang, M. C. K.; Robinson, D. H. Understanding and Learning
Statistics by Computer; World Scientific Publishing Company: Hack-
ensack, NJ , 1986.
9
4496 J. AM. CHEM. SOC. VOL. 132, NO. 12, 2010

Magnetic Glyco-Nanoparticles To Detect Tumor Cells
A R T I C L E S
Differentiation of Closely Related Isogenic Cancer Cells
Including Metastatic Cancer Cells. Besides the differentiation
of cancer vs normal cells, the MR data also enabled the
distinction between closely related isogenic sublines of cancer
cells. Isogenic cancer cells are derived from the same parent
cell line, presenting significant challenges for identification. One
example is the mouse mammary carcinoma cell-lines TA3HA
and TA3ST. These two types of cells originated from the same
parent cell line with TA3HA expressing the mucin-like cell
surface glycoprotein epiglycanin absent in TA3ST.77 Despite
this subtle difference, TA3HA showed significantly stronger
interactions with the Fuc-MGNP 4 and Sia-MGNP 5 (Figure
6b).
Other examples of closely related cells are the widely used
B16F10 and B16F1 mouse melanoma cells, where no qualitative
differences in protein composition, galactose or sialic acid
content on the cell surface, or the membrane fluidity were
observed before.78 The quantitative nature of the MGNP
approach uncovered the subtle difference between these two
cell lines with B16F10 showing larger T2 change (P < 0.00001)
upon binding with Gal-MGNP 3 as compared to B16F1 (Figure
6c), which is likely due to the higher level of galectins expressed
on B16F10 cells.24 This is consistent with the observation that
galactoside mimetics were more potent in preventing the
adhesion of B16F10 to extracellular matrix component than that
of B16F1.24 B16F10 also caused larger ∆T2 with Man-MGNP
2 and Fuc-MGNP 4 at 105 cells/mL. These results indicate that,
despite the overall similarity, there are quantitative changes in
carbohydrate binding between the B16F1 cells and its metastatic
variant B16F10.
Cellular Uptake and Biocompatibility of MGNPs. In order
to gain insights into how MGNPs interact with cells, cellular
staining experiments were performed. Besides their properties
as MRI contrast agents, MGNPs can be visualized by Prussian
blue staining, which yields an intense blue color upon reaction
with the magnetite core of MGNPs, allowing easy tracking of
the particles. As an example, B16F10 cells were incubated with
MGNPs (20 µg/mL), washed extensively with buffer to remove
all unbound nanoparticles, and stained with Prussian blue. As
Man-MGNP 2, Gal-MGNP 3, and Fuc-MGNP 4 caused the
biggest T2 changes when incubated with B16F10, strong blue
stains were observed both on cell surfaces and inside the cells,
suggesting a significant cellular surface binding and cellular
uptake (Figure 7a-c). In contrast, the nonbinding GlcNAc-
MGNP 6 and control particle NP 1 without any carbohydrates
did not show much staining (Figure 7d,e). The same phenomena
were observed with other cell lines as well (Figure S9,
Supporting Information). The correlations between Prussian blue
staining and T2 changes proved that the MR changes were
indeed due to NP binding with the cells. The selective uptake
and intracellular accumulation of specific MGNPs in cancer cells
and the ability of MGNPs to differentiate normal cells from
tumor cells bode well for further development of MGNPs as
vehicles for targeted drug delivery79,80 and magnetic-induced
hyperthermia therapy of cancer.81
Figure 5. LDA plots for the first three LDs of ∆T2 patterns obtained with
the MGNP array upon binding with the 10 cell lines (105 cells/mL). Full
differentiation of the 10 cell lines was achieved.
were then predicted on the basis of the new training set, which
were accurately classified in all the cases tested. This highlights
that, despite the simple structures of the monosaccharides
utilized, the T2 changes of the MGNP array can be employed
as characteristic molecular signatures for each cell line.
Next, the possibility of identifying cell lines with unknown
identity was explored. Two cell lines out of the 10 were
randomly selected as unknowns and submitted to binding with
the six NPs. On the basis of the resulting T2 changes, the
unknowns (unknown 1, MCF-7/Adr-res, and unknown 2, A549)
were classified to the groups previously generated. Both
unknowns were correctly identified, attesting to the reliability
of the analysis (Figure S8, Supporting Information).
Detection of Cancer Cells vs Normal Cells Based on the
MR Signature. A major hurdle for cancer treatment and early
cancer detection is the identification of pertinent cellular
signatures to allow the differentiation of normal cells from their
cancerous counterparts. We envision that this can be achieved
by analysis of the respective cellular characteristics toward
carbohydrate binding. As a proof-of-principle, the T2 changes
of the breast cancer MCF-7/Adr-res cells vs those of the normal
breast endothelial cells 184B5 upon MGNP binding were
examined in detail. As the trend of binding is the same
qualitatively at both cell concentrations, we focused mainly on
T2 changes at 105 cells/mL. The interactions with Man-MGNP
2, Fuc-MGNP 4, and GlcNAc-MGNP 6 were found to be very
similar between the two cell lines (Figure 6a). However, the
MCF-7/Adr-res cells caused a significantly larger ∆T2 upon
binding with Gal-MGNP 3 as compared with that of the normal
breast cells 184B5, which enabled easy detection of breast
cancer cells. This was corroborated by evidence from the
literature that MCF-7/Adr-res cells contain the cancer-specific
galactoside binding galectins-4, -7, and -8, which are absent in
noncancer cell lines.70 Modification of the galactoside ligand
structures76 as well as optimization of ligand density and
nanoparticle surface chemistry can further improve the selectiv-
ity in binding for future in ViVo applications.
(77) Miller, D. K.; Cooper, A. G. J. Biol. Chem. 1978, 253, 8798–8803.
(78) Raz, A.; McLellan, W. L.; Hart, I. R.; Bucana, C. D.; Hoyer, L. C.;
Sela, B. A.; Dragsten, P.; Fidler, I. J. Cancer Res. 1980, 40, 1645–
1651.
(79) Davis, B. G.; Robinson, M. A. Curr. Opin. Drug Disc. DeV. 2002, 5,
279–288.
(76) Delaine, T.; Cumpstey, T.; Ingrassia, L.; Le Mercier, M.; Okechukwu,
P.; Leffler, H.; Kiss, R.; Nilsson, U. J. J. Med. Chem. 2008, 51, 8109–
8114.
(80) Yamazaki, N.; Kojima, S.; Bovin, N. V.; Andre´, S.; Gabius, S.; Gabius,
H.-J. AdV. Drug DeliVery ReV. 2000, 43, 225–244.
(81) Thiesen, B.; Jordan, A. Int. J. Hypertherm. 2008, 24, 467–474.
9
J. AM. CHEM. SOC. VOL. 132, NO. 12, 2010 4497

A R T I C L E S
El-Boubbou et al.
Figure 6. Percentage changes of T2 relaxation time (% ∆T2) obtained upon incubating MGNPs 2-6 or the control NP 1 (20 µg/mL) with (a) breast cancer
MCF-7/Adr-res vs normal breast cell 184B5; (b) TA3ST vs TA3HA cell lines; and (c) B16F1 vs B16F10 (105 cells/mL). Significant differences in binding
with MGNPs were observed, thus differentiating these cell lines.
Figure 7. Prussian blue staining images of 20 µg/mL of (a) Man-MGNP 2; (b) Gal-MGNP 3; (c) Fuc-MGNP 4; (d) GlcNAc-MGNP 6; and (e) control NP
1 incubated with B16F10 cancer cells after unbound particles were removed by washing; and (f) B16F10 cells. The images clearly indicate the high intracellular
uptake of Man-MGNP 2, Gal-MGNP 3, and Fuc-MGNP 4. No Prussian blue stains were visible with the nonbinding GlcNAc-MGNP 6 and the control NP
1, proving the selectivity in binding.
Next, we examined the toxicity of the MGNPs toward the
cells by 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium
bromide (MTT) cell viability assays. Incubation of cells with
different MGNPs for one week did not show any negative
effects on cell viability as compared with untreated cells
(Figure S10, Supporting Information). This demonstrates that
the MGNPs are biocompatible and can be used as safe MRI
contrast agents.
Antiadhesive Properties of MGNPs. Tumor metastasis is
associated with poor prognosis of cancers.82 One of the critical
steps in metastasis is the adhesion of circulating tumor cells to
endothelium at the target location. It has been demonstrated that
cell adhesion inhibitors can be a potential treatment for
metastatic diseases.83-85
which is valuable for guiding the development of antiadhesive
agents, as the strongly binding MGNPs can reduce the
adhesion of cancer cells to the matrix by blocking the cell
surface receptors. According to the MRI signature, Gal-
MGNP 3 interacted strongly with B16F10 cells, which led
us to measure its antiadhesive properties as a proof-of-
principle study. Upon incubation of B16F10 cells with Gal-
MGNP 3, the number of cells adhering to the surface was
reduced by more than 50% (Figure 8). In contrast, the
nonbinding NP 1 showed little effect on cell adhesion.
Modification of the monosaccharide ligands immobilized on
(83) Rojo, J.; Diaz, V.; de la Fuente, J. M.; Segura, I.; Barrientos, A. G.;
Riese, H. H.; Bernad, A.; Penade´s, S. ChemBioChem 2004, 5, 291–
297.
The MRI signature from MGNPs provides detailed infor-
mation on how tumor cells interact with each carbohydrate,
(84) Zhu, D.; Cheng, C.-F.; Pauli, B. U. J. Clin. InVest. 1992, 89, 1718–
1724.
(82) Kawaguchi, T. Current Drug Targets: CardioVasc. Haematol. Dis-
orders 2005, 5, 39–64.
(85) Meromsky, L.; Lotan, R.; Raz, A. Cancer Res. 1986, 46, 5270–5275.
9
4498 J. AM. CHEM. SOC. VOL. 132, NO. 12, 2010

Magnetic Glyco-Nanoparticles To Detect Tumor Cells
A R T I C L E S
receptors, while the wealth of information generated and
magnetic nature of the MGNPs can facilitate future identification
of the receptors. The LDA pattern recognition method was
applied to decipher the glyco-codes of tumor cell binding, which
may be a useful and general protocol to analyze carbohydrate-
receptor interactions.
The strongly binding MGNPs were found to be internalized
by tumor cells, and they significantly reduced the cancer cell
adhesion. As the MGNP array measures the physiologically
related carbohydrate-receptor interactions, which are involved
in a variety of cellular functions including endocytosis, cell-
matrix and cell-cell communications, the knowledge gained
from this new addition to the glyco-nanotechnology toolbox can
enhance our understanding of cancer cell functions. This can
provide leads for further ligand optimization to improve the
specificity in carbohydrate-receptor recognition, which in turn
can enable the application of MGNPs for in ViVo cancer
detection through MRI in the future.
Figure 8. Adhesion of mouse melanoma B16F10 cells (∼5 × 104 cells/
mL) to the surface was significantly reduced by incubation with Gal-MGNP
3 (blue line), while the control NP 1 (green line) had no effect on cell
adhesion as compared to cells without treatment with any NPs (red line).
Error bars indicate standard deviations (triplicate readings).
Acknowledgment. We thank E. Caneva from Centro Interdi-
partimentale Grandi Apparecchiature (C.IGA University of Milan),
L. Polito and S. Ronchi from ISTM-CNR (Milan) for technical
help with HRMAS experiments, and Dr. John Hilkens (Netherlands
Cancer Institute) for kindly providing the two murine mammary
carcinoma cell lines (TA3-HA, TA3-ST). This work was partly
supported by a CAREER award from NSF (X.H.), a predoctoral
fellowship from the American Heart Association (K.B.), and the
Departments of Chemistry and Radiology, Michigan State University.
MGNPs may strengthen the binding with the cells, further
enhancing the antiadhesive effects.
Conclusion
A new approach based on the multichannel MR responses
of MGNPs to qualitatively and quantitatively map the carbo-
hydrate-binding characteristics of a variety of cancer cells was
developed. Validated through binding with a series of lectins
and a well-characterized E. coli system, the carbohydrates
immobilized on MGNPs were found to retain their biological
recognition and binding specificities. Although the monosac-
charides utilized in this study are fairly simple in structures and
multiple cells may bind with the same carbohydrate, the selective
carbohydrate-receptor binding with the MGNP array amplified
the small structural differentials. The resulting combined array
responses allowed the detection of cancer cells as well as the
differentiation of closely related isogenic cancer cell subtypes
without detailed prior knowledge on endogenous carbohydrate
Supporting Information Available: (1) Experimental proce-
dures for synthesis of the MGNPs; (2) characterization of the
MGNPs; (3) MGNP-lectin binding assays; (4) MGNP-E. coli
binding studies; (5) LDA plots at 106 cells/mL; (6) Prussian
blue staining images, and (7) NMR spectra for all new
compounds. This material is available free of charge via the
Internet at http://pubs.acs.org.
JA100455C
9
J. AM. CHEM. SOC. VOL. 132, NO. 12, 2010 4499