Carbohydrate Chips for Studying High-Throughput Carbohydrate-Protein Interactions

Article abstract of DOI:10.1021/ja0391661

Carbohydrate-protein interactions play important biological roles in living organisms. For the most part, biophysical and biochemical methods have been used for studying these biomolecular interactions. Less attention has been given to the development of

Full text of DOI:10.1021/ja0391661

Published on Web 03/26/2004
Carbohydrate Chips for Studying High-Throughput
Carbohydrate-Protein Interactions
Sungjin Park, Myung-ryul Lee, Soon-Jin Pyo, and Injae Shin*
Contribution from the Department of Chemistry, Yonsei UniVersity, Seoul 120-749, Korea
Received October 22, 2003; E-mail: injae@yonsei.ac.kr
Abstract: Carbohydrate-protein interactions play important biological roles in living organisms. For the
most part, biophysical and biochemical methods have been used for studying these biomolecular interactions.
Less attention has been given to the development of high-throughput methods to elucidate recognition
events between carbohydrates and proteins. In the current effort to develop a novel high-throughput tool
for monitoring carbohydrate-protein interactions, we prepared carbohydrate microarrays by immobilizing
maleimide-linked carbohydrates on thiol-derivatized glass slides and carried out lectin binding experiments
by using these microarrays. The results showed that carbohydrates with different structural features
selectively bound to the corresponding lectins with relative binding affinities that correlated with those
obtained from solution-based assays. In addition, binding affinities of lectins to carbohydrates were also
quantitatively analyzed by determining IC50 values of soluble carbohydrates with the carbohydrate
microarrays. To fabricate carbohydrate chips that contained more diverse carbohydrate probes, solution-
phase parallel and enzymatic glycosylations were performed. Three model disaccharides were in parallel
synthesized in solution-phase and used as carbohydrate probes for the fabrication of carbohydrate chips.
Three enzymatic glycosylations on glass slides were consecutively performed to generate carbohydrate
microarrays that contained the complex oligosaccharide, sialyl Lex. Overall, these works demonstrated
that carbohydrate chips could be efficiently prepared by covalent immobilization of maleimide-linked
carbohydrates on the thiol-coated glass slides and applied for the high-throughput analyses of carbohydrate-
protein interactions.
Introduction
of studies using surface plasmon resonance (SPR) spectroscopy
and isothermal titration calorimetry (ITC) have provided
information on the binding affinities of carbohydrates to
proteins.4 In addition, specifically modified synthetic carbohy-
drates have been used to elucidate the molecular basis of
carbohydrate-protein interactions.5
In contrast, much less attention has been given to the
development of high-throughput methods to probe these interac-
tions. During the past decade, DNA chips and protein chips
have been fabricated and utilized in genomic, transcriptomic,
and proteomic investigations. DNA chips have been extensively
used for studying how patterns of gene expression change in
diseases and for simultaneously tracking the activities of many
genes.6 Protein chips, which are more difficult to fabricate than
DNA chips owing to easy denaturation of protein probes, have
been exploited in high-throughput studies of molecular interac-
tions7,8 and in profiling protein expression in normal and
Specific interactions between carbohydrates and proteins
occur in a wide variety of important biological processes,
including cell-cell communication, cell adhesion, fertilization,
differentiation, development, inflammation, and tumor cell
metastasis.1 These interactions also initiate infection of host cells
by bacteria and viruses.1 Therefore, understanding of the
molecular basis for carbohydrate-protein interactions not only
provides valuable information on biological processes in living
organisms but it also aids the development of potent biomedical
agents.2
To date, biophysical and biochemical methods have been
mainly used to probe the details of carbohydrate-protein
interactions. For example, X-ray crystallographic and NMR
spectroscopic techniques have been employed to determine
binding modes between carbohydrates and proteins.3 The results
(1) For recent reviews, see (a) Roth, J. Chem. ReV. 2002, 102, 285. (b) Ritchie,
G. E.; Moffatt, B. E.; Sim, R. B.; Morgan, B. P.; Dwek, R. A.; Rudd, P.
M. Chem. ReV. 2002, 102, 305. (c) Zachara, N. E.; Hart, G. W. Chem.
ReV. 2002, 102, 431. (d) Bertozzi, C. R.; Kiessling, L. L. Science 2001,
291, 2357.
(2) (a) Danishefsky, S. J.; Allen, J. R. A. Angew. Chem., Int. Ed. 2000, 39,
836. (b) Walker, F.; Vicens, Q.; Westhof, E. Curr. Opin. Chem. Biol. 1999,
3, 694. (c) Sears, P.; Wong, C.-H. Angew. Chem., Int. Ed. 1999, 38, 2300.
(3) (a) Mechref, Y.; Novotny, M. V. Chem. ReV. 2002, 102, 321. (b) Wormald,
M. P.; Petrescu, A. J.; Pao, Y.-L.; Glithero, A.; Elliott, T.; Dwek, R. A.
Chem. ReV. 2002, 102, 371. (c) Vyas, N. K. Curr. Opin. Struct. Biol. 1991,
1, 732. (c) Wyss, D. F.; Choi, J. S.; Li, Jing.; Knoppers, M. H.; Willis, K.
J.; Arulanandam, A. R. N.; Smolyar, A.; Reinherz, E. L.; Wagner, G.
Science 1995, 269, 1273.
(4) (a) Dam, T. K.; Brewer, C. F. Chem. ReV. 2002, 102, 387. (b) Green, R.
J.; Frazier, R. A.; Shakesheff, K. M.; Davies, M. C.; Roberts, C. J.; Tendler,
S. J. B. Biomaterials 2000, 21, 1823. (c) Mann, D. A.; Kanai, M.; Maly,
D. J.; Kiessling, L. L. J. Am. Chem. Soc. 1998, 120, 10575.
(5) Lee, Y. C.; Lee, R. T.; Rice, K.; Ichikawa, Y.; Wong, T.-C. Pure Appl.
Chem. 1991, 63, 499.
(6) (a) Perou, C. M. Nature 2000, 406, 747. (b) Ramsey, G. Nat. Biotechnol.
1998, 16, 40. (c) Marshall, A.; Hodgson, J. Nat. Biotechnol. 1998, 16, 27.
(d) DeRisi, J. L.; Lyer, V. R.; Brown, P. O. Science 1997, 278, 680.
(7) (a) Templin, M. F.; Stoll, D.; Schrenk, M.; Traub, P. C.; Vo¨hringer, C. F.;
Joos, T. O. Trends Biotechnol. 2002, 20, 160. (b) Weinberger, S. R.;
Dalmasso, E. A.; Fung, E. T. Curr. Opin. Chem. Biol. 2001, 6, 86.
9
4812
J. AM. CHEM. SOC. 2004, 126, 4812-4819
10.1021/ja0391661 CCC: $27.50 © 2004 American Chemical Society

Carbohydrate Chips
A R T I C L E S
diseased states.9 Much in the same way that DNA chips and
protein chips are used as high-throughput analytic tools,
carbohydrate chips have the potential of serving as useful high-
throughput analytical tools for elucidating recognition events
between carbohydrates and proteins in glycomic researches.
Recent efforts have led to the development of carbohydrate
chips by either covalent or noncovalent immobilization strat-
egy.10 Wang et al. and Feizi et al. successfully fabricated
carbohydrate microarrays by immobilizing various oligosac-
charides on nitrocellulose or nitrocellulose-coated glass slides
without the use of chemical linking techniques.10a,b The im-
mobilization efficiency of carbohydrates was size-dependent;
the smaller carbohydrates were less retained on the nitrocellulose
surface than the larger ones after extensive washing. In addition,
Mrksick et al. fabricated carbohydrate arrays on gold surfaces
via Diels-Alder reactions between cyclopentadiene-possessing
carbohydrates and benzoquinone-coated gold surface.10c The
results indicated that this immobilization method was highly
efficient and oligo(ethylene glycol)-linked carbohydrate mono-
layers displayed minimal nonspecific adsorption with proteins.
We also developed a new method for constructing carbohy-
drate microarrays that relied on covalent immobilization of
carbohydrates on glass slides via chemoselective ligation
between maleimide-linked sugars and thiol-derivatized glass
surface.10d We demonstrated that linker lengths and immobiliza-
tion concentrations of carbohydrates were important factors in
governing protein binding to these microarrays. Importantly,
the results showed that the specificity of protein-carbohydrate
interactions on the solid surfaces resembled that observed in
solution.
In our more recent work in this area, carbohydrate chips
containing more diverse carbohydrate probes have been fabri-
cated to analyze the binding affinities of carbohydrates to several
lectins. Furthermore, three disaccharides were synthesized by
solution-phase parallel glycosylations, and their binding affinities
to lectins on the solid surface were measured and compared
with those of monosaccharides. Finally, a carbohydrate chip
containing the complex oligosaccharide, sialyl Lex, was also
prepared by three consecutive glycosyl transferase-catalyzed
reactions. The results of these studies, which have demonstrated
that carbohydrate microarrays can be efficiently prepared and
used in high-throughput analyses of carbohydrate-protein
interactions, are described below.
Figure 1. Fabrication of carbohydrate microarrays by immobilizing
maleimide-linked carbohydrates on thiol-coated glass slides via chemose-
lective ligation.
Scheme 1. Synthesis of Linkers L1, L2, and L3
(Figure 1).10d For example, carbohydrates linked by L1, L2,
and L3 bound to lectins above 0.5 mM immobilization
concentrations. However, carbohydrates coupled to the shortest
linker S exhibited very weak interactions with lectins below 5
mM immobilization concentrations. Thus, L1, L2, or L3 was
used as a linker for the current studies to fabricate carbohydrate
microarrays containing more diverse carbohydrate probes.
The linkers L1, L2, and L3 were prepared from 6-amino-
hexanoic acid (1) by the procedure shown in Scheme 1. The
amino moiety in 1 was first converted to a maleimide group by
reaction with maleic anhydride followed by treatment with
hexamethyldisilazide (HMDS) and ZnCl2.11,12 The resulting acid
was treated with pentafluorophenol (Pfp-OH), diphenyl chlo-
rophosphate, and N-ethylmorpholine (NEM) to give L1.13 Linker
L2 was prepared by coupling L1 to 1 and a subsequent
esterification with Pfp-OH and N-ethyl-N-(3-dimethylamino-
propyl)carbodiimide‚HCl (EDC). Repetition of this procedure
with L2 provided L3.
The maleimide-linked carbohydrates were prepared by either
a one-pot amination or an allylation approach (Scheme 2). The
one-pot amination of carbohydrates gave â-glycosylamines 2,14
which were coupled to the bifunctional cross-linker L1 or L3
to produce N-linked carbohydrates 3a-g (Figure 2a). Although
this route successfully produced various carbohydrate probes,
this had a limitation to give only â-anomeric carbohydrate
probes.
Results and Discussion
Synthesis of Maleimide-Linked Carbohydrates. We previ-
ously demonstrated that maleimide-linked carbohydrates with
suitable length of tethers on glass slides strongly bound to lectins
(8) (a) MacBeath, G.; Schreiber, S. L. Science 2000, 289, 1760. (b) Zhu, H.;
Klemic, J. F.; Chang, S.; Bertone, P.; Casamayor, A.; Klemic, K. G.; Smith,
D.; Gerstein, M.; Reed, M. A.; Snyder, M. Nat. Genet. 2000, 26, 283.
(9) (a) Merchant, M.; Weinberger, S. R. Electrophoresis 2000, 21, 1164. (b)
Fung, E. T.; Thulasiraman, V.; Weinberger, S. R.; Dalmasso, E. A. Curr.
Opin. Biotech. 2001, 12, 65.
(10) (a) Wang, D.; Liu, S.; Trummer, B. J.; Deng, C.; Wang, A. Nat. Biotech.
2002, 20, 275. (b) Fukui, S.; Feizi, T.; Galustian, C.; Lawson, A. M.; Chai,
W. Nat. Biotech. 2002, 20, 1011. (c) Houseman, B. T.; Mrksich, M. Chem.
Biol. 2002, 9, 443. (d) Park, S.; Shin, I. Angew. Chem., Int. Ed. 2002, 41,
3180. (e) Fazio, F.; Bryan, M. C.; Blixt, O.; Paulson, J. C.; Wong, C.-H.
J. Am. Chem. Soc. 2002, 124, 14397. (f) Willats, W. G. T.; Rasmussen, S.
E.; Kristensen, T.; Mikkelsen, J. D.; Knox, J. P. Proteomics 2002, 2, 1666.
(g) Ko¨hn, M.; Wacker, R. Carsten, P.; Schro¨der, H.; Soule`re, L.; Breinbauer,
R.; Niemeyer, C. M.; Waldmann, H. Angew. Chem., Int. Ed. 2003, 42,
5830. (h) Mellet, C. O.; Ferna´ndez, J. M. G. ChemBioChem. 2002, 3, 819.
(i) Love, K. R.; Seeberger, P. H. Angew. Chem., Int. Ed. 2002, 41, 3583.
(11) Reddy, P. Y.; Kondo, S.; Fujita, S.; Toru, T. Synthesis 1998, 999.
(12) (a) Shin, I.; Jung, H-j.; Lee, M-r. Tetrahedron Lett. 2001, 42, 1325. (b)
Shin, I.; Jung, H.-j.; Cho, J. Bull. Korean Chem. Soc. 2000, 21, 845.
(13) Po¨chlauer, P.; Hendel, W. Tetrahedron 1998, 54, 3489.
(14) (a) Vetter, D.; Gallop, M. A. Bioconjugate Chem. 1995, 6, 316. (b) Vetter,
D.; Tate, E. M.; Gallop, M. A. Bioconjugate Chem. 1995, 6, 319.
9
J. AM. CHEM. SOC. VOL. 126, NO. 15, 2004 4813

A R T I C L E S
Park et al.
Scheme 2. Two Synthetic Routes for the Synthesis of Maleimide-Linked Carbohydrate Probes
Scheme 3. Synthesis of a Maleimide-Linked Sialic Acid Probe
Furthermore, the sialic acid probe 7 was prepared from the pro-
cedure shown in Scheme 3. The known thiosialoside 8 was
transformed to 9 by selective deacetylation of the thioacetate
group with 1 equiv NaOMe and a subsequent coupling of the
resulting thiol to bromoacetylated compound 10.16 Removal of
all the protecting groups in 9 followed by coupling to L2
provided 7.
Fabrication of Carbohydrate Microarrays. Carbohydrate
microarrays (spot size, ca. 100-µm diameter) composed of 22
carbohydrate probes were fabricated by in duplicate printing
solutions of maleimide-connected carbohydrates 3, 6, and 7 (1
nL, 1.0 mM) on a thiol-derivatized glass slide with a high-
precision pin-type microarrayer. Each solution of the carbohy-
drates used in this process was prepared by dissolving the
carbohydrates in phosphate-buffered saline (PBS) containing
40-50% glycerol to suppress undesired evaporation of the
nanodroplets during spotting and immobilization. After 5- or
10-h immobilization in a humid chamber (55% humidity), the
glass slides were treated with 1% N-ethylmaleimide (NEM) in
water to remove unreacted thiol groups. This process prevents
oxidative disulfide-bond formation between surface thiols and
cysteine residues of proteins used in the next incubation step.
A blocked plastic film (0.1-0.2-mm thickness), which was
coated by adhesive at one side, was then carefully attached to
the microspotted slide (Figure 3). Use of the blocked plastic
film facilitates compartmentalization that is required for (1)
simultaneous incubation with several lectins, (2) enzymatic
glycosylation of carbohydrate probes, and (3) determination of
Figure 2. Structures of (a) N-linked and (b) O-linked carbohydrate probes
for the fabrication of carbohydrate microarrays.
Alternatively, the allylation of carbohydrates was performed
to prepare both R-and â-anomeric maleimide-linked carbohy-
drates as well as to examine the difference of binding affinities
between N-linked and O-linked carbohydrates to lectins (Scheme
2b). The R and â-allyl glycosides (4) were reacted with
cysteamine according to the known procedures,15 and the resul-
ting amines (5) were coupled to L2 to yield O-linked R- and
â-anomeric maleimide-linked carbohydrates 6a-g (Figure 2b).
9
4814 J. AM. CHEM. SOC. VOL. 126, NO. 15, 2004

Carbohydrate Chips
A R T I C L E S
Figure 3. A flow chart to fabricate carbohydrate microarrays blocked by
a plastic film.
IC50 values (vide infra). The fabricated carbohydrate microarrays
can be stored in a desiccator for several months without
degradation.
Figure 4. Fluorescence images of carbohydrate microarrays containing 22
carbohydrates probed with (a) FITC-ConA, (b) FITC-NPA, (c) FITC-EC,
(d) FITC-TV, (e) FITC-AA, and (f) fluorescence image of 12000 microspots
(60 × 200) consisting of â-GlcNAc, R-Man, and R-Fuc probed with a
mixture of Cy3-TV, Cy5-AA, and FITC-ConA. The R and â symbols denote
O-linked R and â-anomeric carbohydrates (6) and N denotes N-linked
carbohydrates (3).
Investigations of Protein-Carbohydrate Interactions with
Carbohydrate Microarrays. To profile protein binding with
the carbohydrate microarrays, each block containing 22 carbo-
hydrate probes was treated with 3% bovine serum albumin
(BSA) in PBS containing 0.2% Tween 20 to minimize nonspe-
cific binding of proteins on the surface. Next, the BSA-treated
blocks were probed with solutions of the FITC-labeled lectins
(1-10 µg/mL) such as FITC-ConcanaValin A (FITC-ConA),
FITC-E. cristagalli (FITC-EC), FITC-T. Vulgaris (FITC-TV),
FITC-N. pseudonarcissus (FITC-NPA) and FITC-A. aurantia
(FITC-AA) in PBS containing 0.1% Tween 20 for 0.5-1 h.17,18
After extensive washing of the lectin-treated slide with the same
buffer, lectin binding was visualized and quantitated by using
a fluorescence scanner.
Extensive washing (three times with PBS containing 0.1%
Tween 20 for 5-10 min) of the blocks after incubation with
the FITC-labeled lectins was important for diminishing back-
ground fluorescent signals. The retention of measurable fluo-
rescent signals on carbohydrate probes, even after extensive
washing, suggested that their binding to lectins on the solid
surface was strong. It was well known that carbohydrate-binding
proteins interacted only weakly with monovalent carbohydrates
but they bound to multivalent carbohydrates strongly owing to
cluster effects.19 Thus, carbohydrates immobilized on the glass
slide appear to display multivalency.
Figure 4 shows the fluorescence images of carbohydrate
microarrays probed with five FITC-labeled lectins. The carbo-
hydrate microarray incubated with ConA (an R-Man/R-Glc
binding lectin) exhibited strong binding of R-Man (6e-R), lower
binding of R-Glc (6a-R), and weak binding of maltose (3f) and
R-GlcNAc (6b-R) (Figure 4a).18 These results are consistent
with binding tendencies determined by isothermal titration
calorimetry (ITC).20 In contrast, â-Man (6e-â), â-Glc (3a, 6a-â),
â-GlcNAc (3b, 6b-â), and cellobiose (3g) with â-glucose at
nonreducing terminus were not recognized by ConA. This
demonstrated that the anomeric configuration of the glucose,
mannose, and GlcNAc on the solid surface governed lectin
binding. On the other hand, when the carbohydrate microarray
was treated with NPA, a Man binding lectin which does not
bind to Glc, fluorescent signals were observed only for Man
(17) The lectin of T. Vulgaris (TV) is also known as wheat germ agglutinin
(WGA).
(18) (a) Liener, I. E.; Sharon, N.; Goldstein, I. J. The Lectins: Properties,
Functions, and Applications in Biology and Medicine; Academic Press:
Orlando, FL, 1986; pp 35-244. (b) Van Damme, E. J. M.; Psumans, W.
J.; Pusztai, A.; Bardocz, S. Handbook of Plant Lectins: Properties and
Biomedical Applications; John Wiley & Sons: Chichester, U.K., 1998.
(19) For reviews, see (a) Mannen, M.; Choi, S.-K.; Whitesides, G. M. Angew.
Chem., Int. Ed. 1998, 37, 2754. (b) Lee, Y. C.; Lee, R. T. Acc. Chem. Res.
1995, 28, 321.
(15) (a) Lee, R. T.; Lee, Y. C. Carbohydr. Res. 1974, 37, 193. (b) Wong, T. C.;
Townsend, T. R.; Lee, Y. C. Carbohydr. Res. 1987, 170, 27. (c) Utille,
J.-P.; Priem, B. Carbohydr. Res. 2000, 329, 431. (d) Garegg, P. J.; Norberg,
T. Carbohydr. Res. 1976, 52, 235. (e) Rosenberg, H. J.; Riley, A. M.;
Marwood, R. D.; Correa, V.; Taylor, C. W.; Potter, B. V. L. Carbohydr.
Res. 2001, 332, 53.
(20) Dissociation constants (Kd) of methyl R-mannoside, methyl R-glucoside,
maltose, methyl R-GlcNAc, and ManR1,6Man-R-OMe for ConA deter-
mined by ITC were 0.12, 0.51, 0.76, 0.93, and 0.12 mM, respectively.
Mandal, D. K.; Kishore, N.; Brewer, C. F. Biochemistry 1994, 33, 1149.
(16) Fitz, W.; Rosenthal, P. B.; Wong, C.-H. Bioorg. Med. Chem. 1996, 4, 1349.
9
J. AM. CHEM. SOC. VOL. 126, NO. 15, 2004 4815

A R T I C L E S
Park et al.
(6e-R and â) and the intensity of microspots of R-Man (6e-R)
was greater than that of â-Man (6e-â) (Figure 4b).21 The relative
fluorescence intensity in the region of R-Man (6e-R) for ConA
and NPA showed that ConA bound to R-Man more strongly
than NPA. The dissociation constants (Kd) of ManR1,3Man for
ConA and NPA determined by ITC were reported to be 0.07
and 2.00 mM, respectively.20,22 Thus, binding experiments using
both the carbohydrate chips and the ITC method indicated that
ConA interacted more strongly with Man than NPA.
The carbohydrate microarray incubated with EC (a GalNAc/
Gal binding lectin) showed fluorescent signals in the regions
of Gal, GalNAc, and lactose (Figure 4c).18 The R-GalNAc
(6d-R), N-linked â-GalNAc (3c), and lactose (3e) bound to EC
with similar binding affinities. However, O-linked Gal (6c-R
and â) and â-GalNAc (6d-â) interacted with the lectin with
the reduced binding affinities. Interestingly, N-linked â-GalNAc
(3c) exhibited stronger fluorescence intensity than O-linked
â-GalNAc (6d-â). This suggested that the nature of the anomeric
linkage in maleimide-linked monosaccharides affected their
binding affinities to the lectin. The relative binding affinities
of EC with the carbohydrate microarray were similar to those
observed for solution-based assays.23
Figure 5. Determination of concentrations (IC50) of soluble methyl
N-acetyl-R-glucosaminide (R-GlcNAc-OMe) to inhibit 50% of TV binding
to R-GlcNAc (a) and R-GalNAc (b) on the carbohydrate microarray, and
determination of concentrations (IC50) of soluble allyl R-fucoside (c) and
methyl R-fucoside (d) to inhibit 50% of AA binding to â-Man on the
carbohydrate microarray.
When the carbohydrate microarray was probed with TV (a
GlcNAc binding lectin), the lectin recognized R-GlcNAc (6b-R)
strongly and O-linked â-GlcNAc (6b-â) with the reduced
binding affinity (Figure 4d).18 Microspots of N-linked â-GlcNAc
(3b) exhibited weaker fluorescent signals than those of O-linked
â-GlcNAc (6b-â). It was also revealed that TV bound to
O-linked R-GalNAc (6d-R) and O-linked â-GlcNAc (6b-â) with
similar binding affinities but to â-GalNAc (3c, 6d-â) with
reduced binding affinity.18,24 To analyze the difference of
binding affinity between R-GlcNAc (6b-R) and R-GalNAc
(6d-R) for TV, we determined concentrations (IC50) of methyl
N-acetyl-R-glucosaminide (R-GlcNAc-OMe) to inhibit 50% of
TV binding to R-GlcNAc (6b-R) and R-GalNAc (6d-R) on the
microarray. Microspots of R-GlcNAc and R-GalNAc were
incubated with a series of mixtures of Cy3-TV (1 µg/mL) and
R-GlcNAc-OMe (∼10 µM to 0.4 M) for 1 h. After extensive
washing of the carbohydrate chip, the amount of bound lectin
was quantitated by measuring fluorescence intensity. As shown
in Figure 5a and b, R-GlcNAc (IC50 ) 8.9 mM for R-GlcNAc-
OMe) competed more efficiently with R-GlcNAc-OMe for TV
than R-GalNAc (IC50 ) 3.5 mM for R-GlcNAc-OMe).25
Because of configurational similarity of GlcNAc and N-
acetylneuraminic acid (NeuNAc) at positions C-2 (NHAc) and
C-3 (OH), TV was known to recognize terminal NeuNAc.24b
However, it was revealed that monomeric thiosialoside 7 on
the glass slide did not bind to the lectin.
â) but weak binding of N-linked â-Fuc (3d) (Figure 4e).18
Microspots of N-linked â-Fuc (3d) displayed weaker fluorescent
signals than those of O-linked â-Fuc (6f-â). It was intriguing
that AA also bound to O-linked â-Man (6e-â) with a similar
affinity as it did to N-linked â-Fuc (3d). Since this was an
unexpected result, the binding affinity of O-linked â-Man (6e-â)
to AA was assessed by determining concentrations (IC50) of
methyl R-fucoside (R-Fuc-OMe) and allyl R-fucoside (R-Fuc-
OAllyl) that inhibit 50% of AA binding to the O-linked â-Man
(6e-â).26 For this purpose, microspots of O-linked â-Man (6e-â)
were incubated with a series of mixtures of Cy5-AA (1 µg/
mL) and R-Fuc-OMe or R-Fuc-OAllyl (1 nM-0.01 M) for 1 h,
and then the amount of bound lectin was quantitated by meas-
uring fluorescence intensity. In this manner, the IC50 values of
R-Fuc-OAllyl and R-Fuc-OMe were determined to be 6.2 and
3.5 µM, respectively (Figure 5c and d). The results suggested that
O-linked â-Man (6e-â) was a poor ligand for AA and R-Fuc-
OMe was a little better ligand for the lectin than R-Fuc-OAllyl.
To demonstrate that carbohydrates on the chips containing
several microspots selectively bind to proteins, we printed 12000
microspots (60 × 200) composed of R-Fuc (6f-R), R-Man
(6e-R), and O-linked â-GlcNAc (6b-â) on a thiol-coated slide
in an alternate fashion. The microarray was then probed with a
mixture of Cy3-TV, Cy5-AA, and FITC-ConA. As shown in
Figure 4f, R-Fuc, R-Man, and â-GlcNAc selectively bound to
the corresponding lectins. The lectin-binding experiments
described above clearly demonstrated that carbohydrate chips
were ideally suited for high-throughput analysis of specific
carbohydrate-protein interactions.
(24) (a) Monsigny, M.; Roche, A.-C.; Sene, C.; Maget-Dana, R.; Delmotte, F.
Eur. J. Biochem. 1980, 104, 147. (b) Wright, C. S. J. Mol. Biol. 1980,
141, 267. (c) Bhavanandan, V. P.; Kaltlic, A. W. J. Biol. Chem. 1979,
254, 4000.
The carbohydrate microarray probed with FITC-AA (a Fuc
binding lectin) showed strong binding of O-linked Fuc (6f-R and
(25) Goldstein, I. J.; Hammarstro¨m, S.; Sundblad, G. Biochim. Biophys. Acta
1975, 405, 63.
(21) Kaku, H.; Van Damme, E. J. M.; Peumans, W. J.; Goldstein, I. J. Arch.
Biochem. Biophys. 1990, 279, 298.
(26) It was reported that AA did not bind to Man up to 50 mM concentration.
However, it seems that AA weakly binds to Man on the solid surface maybe
because of cluster effect. (a) Kochibe, N.; Furukawa, K. Biochemistry 1980,
19, 2841. (b) Sudakevitz, D.; Imberty, A.; Gilboa-Garber, N. J. Biochem.
2002, 132, 353.
(22) Sauerborn, M. K.; Wright, L. M.; Reynolds, C. D.; Grossmann, J. G.;
Rizkallah, P. J. J. Mol. Biol. 1999, 290, 185.
(23) Kaladas, P. M.; Kabat, E. A.; Iglesias, J. L.; Lis, H.; Sharon, N. Arch.
Biochem. Biophys. 1982, 217, 624.
9
4816 J. AM. CHEM. SOC. VOL. 126, NO. 15, 2004

Carbohydrate Chips
A R T I C L E S
Scheme 4. Solution-Phase Parallel Synthesis of Three
Disaccharides. Conditions: (i) HSCH2CH2NHBoc, TEA, DMF, 50
°C; (ii) NaOMe, MeOH; (iii) TFA; (iv) L2, DIEA, DMF
Figure 6. Fluorescence images of microspots containing mono- and
disaccharides probed with FITC-labeled lectins.
labeled lectins (ConA, NPA, EC, TV, and AA). As shown in
Figure 6, ConA bound to ManR1,6Man (20) and O-linked
R-Man (6e-R) with similar binding affinities.20 However, NPA
interacted with ManR1,6Man (20) more strongly than it did with
O-linked R-Man (6e-R). Microspots treated with EC exhibited
stronger fluorescent signals in the region of O-linked â-GalNAc
(6d-â) than Galâ1,6Man (19). As anticipated, TV and AA only
bound to the monosaccharide controls.
Enzymatic Transformations of Carbohydrates on the
Carbohydrate Microarrays. Enzymatic glycosylation pro-
cesses serve as alternatives to chemical glycosylations for the
synthesis of complex carbohydrates.27f,29 To examine the
usefulness of enzymatic transformations of carbohydrates on
carbohydrate microarrays, we prepared sialyl Lex from GlcNAc
by three consecutive glycosyl transferase-catalyzed reactions.
The N-linked â-GlcNAc (3b) was in quadruplicate microspot-
ted on the thiol-coated glass slide and then compartmentalized
by using a plastic film to create separate blocks. Efficient
immobilization of â-GlcNAc was confirmed by probing one
block with Cy3-TV (Figure 7a). The remaining blocks were
treated with â-1,4-galactosyl transferase (GalT) in the presence
of UDP-Gal for 15 h at 37 °C in a humid chamber.10c To prove
successful enzymatic galactosylation, one block was incubated
with Cy5-EC. High fluorescent signals were observed in the
region of microspots containing LacNAc. The GalT-incubated
block was further treated with R-2,3-sialyl transferase (SialT)
and CMP-NeuAc under the above-mentioned conditions.29 The
resulting microarray was then probed with Cy5-EC. No
fluorescence was observed, indicating that LacNAc was con-
verted to NeuNAcR2,3LacNAc.
Finally, a solution of R-1,3-fucosyl transferase (FucT) and
GDP-Fuc was added to a block containing NeuNAcR2,-
3LacNAc.10e After 15 h at 37 °C in a humid chamber, the glass
slide was rinsed with PBS three times, and then the enzymatic
fucosylation was repeated under the same conditions to bring
about more fucosylation of NeuNAcR2,3LacNAc.30 The block
was then treated sequentially with mouse anti-sialyl Lex and
goat Cy5-anti-antibody. As shown in Figure 7a, microspots of
the synthetic sialyl Lex displayed visible fluorescence. A single
fucosylation leads to poor glycosylation on the basis of low
fluorescence intensity.
Preparation of Maleimide-Linked Carbohydrate Probes
by Solution-Phase Parallel Glycosylation. Carbohydrate mi-
croarrays composed of various carbohydrate probes are poten-
tially more useful for understanding recognition events between
carbohydrates and proteins. Recently, combinatorial synthesis
of carbohydrates have been developed to prepare diverse
carbohydrate libraries.27 In the current effort, we synthesized
three model maleimide-linked disaccharides (ManR1,6Man,
Glcâ1,6Man, and Galâ1,6Man) by solution-phase parallel
glycosylation (Scheme 4). The 2-bromoethyl R-mannoside 11
was glycosylated with three glycosyl bromides 12-14 in the
presence of silver triflate (AgOTf) as an activator to give the
1,6-linked disaccharides 15-17.27c The bromoethyl group was
used as an anomeric substituent because it could be readily
transformed into other functional groups.27c,28 The three disac-
charides (15-17) were converted to maleimide-linked carbo-
hydrates 18-20 by reaction with Boc-protected cysteamine,
removal of all the protecting groups, and a subsequent coupling
of the resulting amines to L2.
The three disaccharides were printed in duplicate on a thiol-
coated slide along with four monosaccharides (6b-R, 6d-â, 6e-R,
and 6f-â) as controls and then were probed with five FITC-
The selective glycosylations of carbohydrates by glycosyl
transferases on carbohydrate microarrays was also examined.
For this purpose, GlcNAc (3b) and R-Fuc (6f-R) were alternately
printed on a glass slide to produce 100 microspots (5 × 20).
The slide was then incubated with GalT and UDP-Gal under
(27) (a) Marcaurelle, L. A.; Seeberger, P. H. Curr. Opin. Chem. Biol. 2002, 6,
289. (b) Seeberger, P. H.; Haase, W.-C. Chem. ReV. 2000, 100, 4349. (c)
Takahashi, T.; Adachi, M.; Matsuda, A.; Doi, T. Tetrahedron Lett. 2000,
41, 2599. (d) Sofia, M. J. Mol. DiVers. 1998, 3, 75. (e) Liang, R.; Yan, L.;
Loebach, J.; Ge, M.; Uozumi, Y.; Sekanina, K.; Horan, N.; Gildersleeve,
J.; Thompson, C.; Smith, A.; Biswas, K.; Still, W. C.; Kahne, D. Science
1996, 274, 1520. (f) Koeller, K. M.; Wong, C.-H. Chem. ReV. 2000, 100,
4465.
(29) Gijsen, H. J. M.; Qiao, L.; Fitz, W.; Wong, C.-H. Chem. ReV. 1996, 96,
(28) Kieburg, C.; Sadalapure, K.; Lindhorst, T. K. Eur. J. Org. Chem. 2000,
2035.
443.
(30) Seitz, O.; Wong, C.-H. J. Am. Chem. Soc. 1997, 119, 8766.
9
J. AM. CHEM. SOC. VOL. 126, NO. 15, 2004 4817

A R T I C L E S
Park et al.
processing enzymes, such as glycosyl transferases, glycosidases,
and sulfotransferases. Finally, carbohydrate microarrays might
play a potential role in the diagnoses of diseases which correlate
with the presence or absence of carbohydrate-binding molecules.
Experimental Section
General. All materials were purchased from commercial suppliers:
solvents and reagents from Aldrich, Acros, and Senn Chemicals; thiol-
derivatized glass slides from Biometrix Technology; lectins from Vector
labs and Sigma; glycosyl transferases, substrates, and antibodies from
Calbiochem; Cy3 and Cy5 N-hydroxysuccinimide ester from Amersham
Pharmacia biotech. The solutions of carbohydrates were microspotted
with a MicroSys 5100 from Cartesian Technologies. Carbohydrate
microarrays probed with fluorescent dye-labeled lectins were scanned
with a Scanarray 5000 from Packardbiochip.
Microspotting of Carbohydrates. The carbohydrate probes were
dissolved in PBS (pH 6.8) containing 40-50% glycerol. The solution
of carbohydrates (1 nL, 1.0 mM) from a 384-well plate was microspot-
ted in predetermined places on a thiol-coated glass slide with a distance
of 200 µm between the centers of adjacent spots (spot size, ca. 100
µm in diameter). After completion of printing, the slide was placed in
a humid chamber (55% humidity) at room temperature for 5 or 10 h
and then immersed into water containing 1% N-ethylmaleimide (NEM)
for 15 min with gentle shaking. The slide was washed with PBS (pH
6.8) containing 0.1% Tween 20 for 1 h and then rinsed with water.
After drying by purging with Ar gas, a compartmentalized plastic film
which was coated by adhesive at one side (thickness: 0.1-0.2 mm)
was attached to the glass slide. The solution of 0.2% Tween 20
possessing 3% BSA was dropped in the blocks and then incubated for
30 min. The slide was washed with the same buffer without BSA (3 ×
15 min).
Figure 7. (a) Enzymatic synthesis of sialyl Lex from GlcNAc by three
glycosyl transferases. (b) Selective glycosylation of GlcNAc by galactosyl
transferase.
Detection of Lectin-Carbohydrate Interactions. Blocks contain-
ing microspots were probed with fluorescent dye-labeled lectins (1-
10 µg/mL) in PBS (pH 6.8) containing 0.1% Tween 20 for 1 h at room
temperature. For ConA binding, MnCl2 and CaCl2 were added at final
concentrations of 1 mM. The unbound lectins were then removed by
gentle shaking in the same buffer (3 × 5-10 min). After removal of
a plastic film, the slide was scanned by using a Scanarray 5000.
Quantitative Analysis of Lectin Binding to Microspots. The
carbohydrate probes were microspotted in predetermined places on the
glass slide. After attaching a compartmentalized plastic film to the
microspotted slide, each block was incubated with a series of mixtures
of fluorescent dye-labeled lectin and an inhibitor in PBS (pH 6.8)
containing 0.1% Tween 20 for 1 h. Fluorescence intensity was measured
by using a Scanarray 5000.
the glycosylation conditions given above. The results of probing
the carbohydrate chip with a mixture of Cy5-AA and FITC-EC
show that only GlcNAc was converted to LacNAc by GalT
(Figure 7b). These observations showed that carbohydrate probes
were efficiently and selectively transformed by glycosyl trans-
ferases to produce carbohydrate microarrays that contained more
complex and perhaps more diverse carbohydrates.
Conclusion
In the studies described above, we demonstrated that carbo-
hydrate microarrays were suitable for sensitive, high-throughput
detection of binding between carbohydrate ligands and proteins.
We found that the nature of the anomeric linkage in maleimide-
linked monomeric carbohydrates affected their protein binding
affinities. In addition, we showed that the carbohydrate mi-
croarrays could be used to quantitatively analyze binding
affinities between carbohydrates and lectins. We also showed
that solution-phase parallel glycosylations could be employed
to prepare carbohydrate microarrays that contained diverse
probes. Finally, we demonstrated that enzymatic glycosylations
on the glass slides could be applied for the synthesis of complex
oligosaccharides such as sialyl Lex.
Enzymatic Synthesis of Sialyl Lex on the Carbohydrate Microar-
rays. The slide microspotted by GlcNAc (3b, 1 mM) in predetermined
places was compartmentalized with a plastic film. A solution (15 µL)
of GalT (23 mU), MnCl2 (10 mM), and UDP-Gal (0.1 mM) in HEPES
buffer (50 mM, pH 7.5) was dropped into each block and the slide
was placed into a humid chamber for 15 h at 37 °C. The slide was
washed with PBS (pH 6.8) containing 0.1% Tween 20 for 3 × 10 min
and then dried by purging with Ar gas. For sialylation of LacNAc, a
solution (15 µL) of SialT (1 mU), MnCl2 (5 mM), alkaline phosphatase
(20 µU), and CMP-NeuNAc (0.1 mM) in HEPES buffer (100 mM, pH
7.0) was dropped in each block and then placed in the humid chamber
for 15 h at 37 °C. The slide was washed with the above-mentioned
buffer. For fucosylation of NeuNAcR2,6LacNAc, a solution (15 µL)
of FucT (1 mU), MnCl2 (15 mM), alkaline phosphatase (20 µU), and
GDP-Fuc (0.1 mM) in MES buffer (50 mM, pH 6.0) was dropped in
each block and then placed in the humid chamber for 15 h at 37 °C.
The slide was washed with the same buffer.
In addition to their uses in high-throughput profiling of protein
binding, carbohydrate microarrays might be applicable to high-
throughput screening of potential inhibitors for carbohydrate-
binding proteins required for the development of novel thera-
peutic agents. Moreover, they might be used to identify or
characterize novel carbohydrate-binding proteins or carbohydrate-
To examine successful sialylation, one block after incubation with
SialT was treated with 0.2% Tween 20 containing 3% BSA for 30
min and then probed with Cy5-EC (10 µg/mL). To test efficient
9
4818 J. AM. CHEM. SOC. VOL. 126, NO. 15, 2004

Carbohydrate Chips
A R T I C L E S
fucosylation, after incubation with FucT one block was incubated with
anti-sialyl Lex antibody (5 µg/mL) in PBS (pH 7.2) containing NaCl
(500 mM) and 0.02% Tween 20 for 1 h at room temperature and then
was probed with Cy5-anti-antibody (10 µg/mL) in the same buffer for
1.5 h at room temperature.
with a solution (50 µL) of FITC-EC and Cy5-AA in PBS (pH 6.8)
containing 0.1% Tween 20 for 1 h at room temperature.
Acknowledgment. This work was supported by Ministry of
Science & Technology (M10213050001-02B1505-00210).
Selective Enzymatic Glycosylation on the Glass Slide. The
solutions of GlcNAc (3b, 1 mM) and Fuc (6f-R, 1 mM) were alternately
microspotted on the thiol-derivatized glass slide. The carbohydrate
microarrays were incubated with a solution (15 µL) of GalT (23 mU),
MnCl2 (10 mM), and UDP-Gal (0.1 mM) in HEPES buffer (50 mM,
pH 7.5) in the humid chamber for 15 h at 37 °C. The slide was probed
Supporting Information Available: Detail synthetic proce-
dures and NMR spectra for carbohydrate probes (PDF). This
material is available free of charge via the Internet at
http://pubs.acs.org.
JA0391661
9
J. AM. CHEM. SOC. VOL. 126, NO. 15, 2004 4819