Surface plasmon resonance imaging studies of protein-carbohydrate interactions

Article abstract of DOI:10.1021/ja034165u

Carbohydrate arrays fabricated on gold films were used to study carbohydrate-protein interactions with surface plasmon resonance (SPR) imaging. An immobilization scheme consisting of the formation of a surface disulfide bond was used to attach thiol-modified carbohydrates onto gold films and to fabricate carbohydrate arrays. The carbohydrate attachment steps were characterized using polarization modulation Fourier transform infrared reflection absorption spectroscopy; and poly(dimethylsiloxane) microchannels were used to immobilize probe compounds at discrete locations on a gold film. The binding of the carbohydrate-binding proteins concanavalin A (ConA) and jacalin to arrays composed of the monosaccharides mannose and galactose was monitored with SPR imaging. SPR imaging measurements were employed to accomplish the following: (i) construct adsorption isotherms for the interactions of ConA and jacalin to the carbohydrate surfaces, (ii) monitor protein binding to surfaces presenting different compositions of the immobilized carbohydrates, and (iii) measure the solution equilibrium dissociation constants for ConA and jacalin toward mannose and galactose, respectively. Adsorption coefficients (K_ADS) of 2.2 ± 0.8 × 10⁷ M^-1 and 5.6 ± 1.7 × 10⁶ M^-1 were obtained for jacalin adsorbing to a galactose surface and ConA adsorbing to a mannose surface, respectively. The solution equilibrium dissociation (K_D) constant for the interaction of jacalin and galactose was found to be 16 ± 5 μM, and for ConA and mannose was found to be 200 ± 50 μM.

Full text of DOI:10.1021/ja034165u

Published on Web 04/26/2003
Surface Plasmon Resonance Imaging Studies of
Protein-Carbohydrate Interactions
Emily A. Smith, William D. Thomas, Laura L. Kiessling,* and Robert M. Corn*
Contribution from the Department of Chemistry, UniVersity of Wisconsin, 1101 UniVersity AVe.,
Madison, Wisconsin 53706, USA
Received January 14, 2003; E-mail: corn@chem.wisc.edu; kiessling@chem.wisc.edu
Abstract: Carbohydrate arrays fabricated on gold films were used to study carbohydrate-protein interactions
with surface plasmon resonance (SPR) imaging. An immobilization scheme consisting of the formation of
a surface disulfide bond was used to attach thiol-modified carbohydrates onto gold films and to fabricate
carbohydrate arrays. The carbohydrate attachment steps were characterized using polarization modulation
Fourier transform infrared reflection absorption spectroscopy; and poly(dimethylsiloxane) microchannels
were used to immobilize probe compounds at discrete locations on a gold film. The binding of the
carbohydrate-binding proteins concanavalin A (ConA) and jacalin to arrays composed of the monosac-
charides mannose and galactose was monitored with SPR imaging. SPR imaging measurements were
employed to accomplish the following: (i) construct adsorption isotherms for the interactions of ConA and
jacalin to the carbohydrate surfaces, (ii) monitor protein binding to surfaces presenting different compositions
of the immobilized carbohydrates, and (iii) measure the solution equilibrium dissociation constants for ConA
and jacalin toward mannose and galactose, respectively. Adsorption coefficients (KADS) of 2.2 ( 0.8 × 107
M-1 and 5.6 ( 1.7 × 106 M-1 were obtained for jacalin adsorbing to a galactose surface and ConA adsorbing
to a mannose surface, respectively. The solution equilibrium dissociation (KD) constant for the interaction
of jacalin and galactose was found to be 16 ( 5 µM, and for ConA and mannose was found to be 200 (
50 µM.
Introduction
interactions using solution-based analysis methods such as those
traditionally used to study protein or nucleic acid interactions.8
Protein-carbohydrate interactions underlie many aspects of
cellular recognition including cell adhesion, trafficking, apo-
ptosis, and the immune response.1-3 These specific interactions
occur through glycoprotein, glycolipid, and polysaccharide
displays found on cell surfaces and lectins, proteins with
carbohydrate-binding domains.4 Lectins typically possess shal-
low binding pockets that are solvent-exposed; therefore, the
interaction between the carbohydrate and the protein is typically
weak.5-7 To provide interaction strength and specificity, many
carbohydrate-binding proteins are oligomeric, consisting of
several similar or identical monomers that each bind to a
carbohydrate.1 It is possible for the lectins to participate in
multivalent binding, or the formation of several simultaneous
binding events that provide an apparent binding affinity
(functional affinity) that is greater than the sum of the individual
interactions.
Surface-based carbohydrate arrays can be used to facilitate the
study of lectin recognition, and these will be valuable tools in
the field of glycomics.9-13 The presentation of carbohydrates
in an array format provides a means to simultaneously monitor
multiple binding events between immobilized carbohydrates and
proteins binding from solution. In addition, carbohydrates
immobilized on a surface are presented multivalently.14 Lectin
binding to a multivalent array is more avid and of higher
specificity than the interactions of the monovalent counter-
parts.14-19 Thus, immobilized arrays of ligands facilitate the
analysis of lectin binding and can be relevant to cell surface
presentations. To date, there are few reports9-13 of the imple-
(8) Mann, D. A.; Kanai, M.; Maly, D. J.; Kiessling, L. L. J. Am. Chem. Soc.
1998, 120, 10 575-10 582.
(9) Kiessling, L. L.; Cairo, C. W. Nature Biotech. 2002, 20, 234-235.
(10) Wang, D.; Liu, S.; Trummer, B. J.; Deng, C.; Wang, A. Nature Biotech.
2002, 20, 275-281.
Tools for studying protein-carbohydrate interactions are
necessary to gain an understanding of biological function and
the roles these interactions play in disease states. Monovalent
protein-carbohydrate interactions in solution are often of low
affinity, which makes it difficult to analyze the binding
(11) Fukui, S.; Feizi, T.; Galustian, C.; Lawson, A. M.; Chai, W. Nature Biotech.
2002, 20, 1011-1017.
(12) Houseman, B. T.; Mrksich, M. Chem. Biol. 2002, 9, 443-454.
(13) Fazio, F.; Bryan, M. C.; Blixt, O.; Paulson, J. C.; Wong, C.-H. J. Am.
Chem. Soc. 2002, 124, 14 397-14 402.
(14) Houseman, B. T.; Mrksich, M. Top. Curr. Chem. 2002, 218, 1-44.
(15) Mortell, K. H.; Weatherman, R. V.; Kiessling, L. L. J. Am. Chem. Soc.
1996, 118, 2297-2298.
(1) Lis, H.; Sharon, N. Chem. ReV. 1998, 98, 637-674.
(2) Varki, A. Glycobiology 1993, 3, 97-130.
(16) Lee, R. T.; Lee, Y. C. Glycoconjugate J. 1987, 4, 317.
(17) 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, 1520-1522.
(3) Bertozzi, C. R.; Kiessling, L. L. Science 2001, 291, 2357-2364.
(4) Sharon, N.; Lis, H. Science 1989, 246, 227-246.
(5) Kiessling, L. L.; Pohl, N. L. Chem. Biol. 1996, 3, 71-77.
(6) Rini, J. M. Curr. Opin. Struct. Biol. 1995, 5, 617-621.
(7) Weis, W. I.; Drickamer, K. Annu. ReV. Biochem. 1996, 65, 441-473.
(18) Liang, R.; Loebach, J.; Horan, N.; Ge, M.; Thompson, C.; Yan, L.; Kahne,
D. Proc. Natl. Acad. Sci. U.S.A. 1997, 94, 10 554-10 559.
(19) Lee, Y. C.; Lee, R. T. Acc. Chem. Res. 1995, 28, 321-327.
9
6140
J. AM. CHEM. SOC. 2003, 125, 6140-6148
10.1021/ja034165u CCC: $25.00 © 2003 American Chemical Society

SPR Imaging Studies of Protein-Carbohydrate Interactions
A R T I C L E S
for the synthesis of compounds 1 and 2 are provided in the Supporting
Information. Other standard chemicals were purchased from commercial
suppliers, and used as received.
mentation of carbohydrate arrays due to the lack of robust,
general, and controlled array fabrication strategies for carbo-
hydrates.
Disulfide Attachment Chemistry. The MUAM monolayer was
prepared by soaking a gold-coated slide in a 1.0 mM MUAM in ethanol
solution for at least 24 h. The reaction of the amine-substituted surface
with the succinimide ester of SATP was carried out by treating the
surface with 4.0 mM SATP in a mixture of dimethylformamide (DMF)
and 0.1 M triethanolamine (10:90 DMF/triethanolamine) solution (pH
7.0) for 1-2 h. The thioester was cleaved by soaking the slide for 20
min in a solution of 0.5 M hydroxylamine, 0.05 M dithiothreitol (DTT),
0.05 M phosphate buffer and 0.025 M EDTA at pH 7.5 to generate the
free thiol. To generate the intermediate mixed disulfide, the thiol-
substituted surface was treated with 2,2′-dipyridyl disulfide (1 mg/mL)
in a 1:1 mixture of 0.1 M triethanolamine (pH 8.0) and DMF for 2 h.
Immobilization of the carbohydrate derivatives, compounds 1 and 2,
was carried out in 20 mM phosphate buffer (pH 7.5, 100 mM NaCl, 5
mM MgCl2) for 12 h. Between each reaction step the slides were
thoroughly rinsed with water and dried under a nitrogen stream.
In addition to the fabrication of carbohydrate arrays, a method
to analyze protein binding to these arrays that does not require
the use of fluorescent, radioactive, or enzymatic reporter groups
is desirable. It is also desirable to circumvent the use of a
secondary binding event for detection. We sought to employ
surface plasmon resonance (SPR) imaging, an optical technique
that is used to spatially monitor localized differences in the
reflectivity of incident light from a prism-gold film interface
that result from molecules adsorbing to or desorbing from the
gold film. SPR imaging can be used to directly study the
interactions of the carbohydrate arrays with proteins adsorbing
from solution without the use of a reporter group.20-23 In
addition to the benefits afforded from the increased functional
affinity and specificity of a surface-based assay format, the use
of SPR to monitor protein-carbohydrate interactions with
carbohydrate arrays has distinct advantages over fluorescence
methods. Specifically, lower affinity interactions can be detected
because SPR measurements can be made in the presence of a
large excess of unbound protein, without the background
problems that would complicate a similar measurement using
fluorescence based techniques.
In this paper, we report the fabrication of mannose and
galactose carbohydrate arrays on gold films using poly-
(dimethylsiloxane) (PDMS) microchannels, and their subsequent
use in SPR imaging experiments to monitor the binding of the
lectins jacalin and concanavalin A (ConA). The attachment of
the mannose and galactose ligands to the surface was character-
ized with polarization modulation Fourier transform infrared
reflection absorption spectroscopy (PM-FTIRRAS). SPR imag-
ing measurements were used to demonstrate that the im-
mobilized carbohydrates are accessible to proteins in solution
and that the lectin binding specificities could be detected.
Adsorption isotherms for the binding of ConA and jacalin to
carbohydrate surfaces were constructed to determine the adsorp-
tion coefficients (KADS) for jacalin and ConA to the carbohydrate-
substituted surfaces, and the binding of protein to surfaces
presenting different carbohydrate compositions was studied.
Finally, SPR imaging measurements were used to determine
solution equilibrium dissociation constants (KD) using a model
that was derived from the multivalent Scatchard analysis24 and
competition binding experiments, in which carbohydrate was
present in solution with the protein.
PM-FTIRRAS Surface Characterization. Commercial gold slides
(5 nm Cr and 100 nm Au) were purchased from Evaporated Metal
Films (New York) and were used for all PM-FTIRRAS measurements.
Mid-IR spectra were collected using a Mattson RS-1 spectrometer with
real-time interferogram sampling electronics and optical layout, as
previously described.25,26 Spectra were collected from 1000 scans with
a resolution of 4 cm-1 using a narrow-band HgCdTe detector. Spectra
from the CH stretching region were collected using a Bruker Vector-
22 spectrometer with the same sampling electronics and optical layout
as described above. Spectra were collected from 1000 scans with a
resolution of 2 cm-1 using a InSb detector.
Surface Plasmon Resonance Imaging. Gold films (45 nm) with a
thin chromium underlayer (1 nm) used to construct the arrays were
vapor-deposited on SF10 glass slides (Schott Glass Technologies) in a
Denton Vacuum DV-502A evaporator. The in situ SPR imaging
apparatus has been described elsewhere.22,27,28 Briefly, p-polarized
collimated white light was directed toward a prism/Au thin film/buffer
assembly at a fixed angle. Light reflected from this assembly was passed
through a narrow band-pass filter (λ ) 800 nm) and collected by a
CCD camera. The carbohydrate arrays were imaged in 25 mM tris-
(hydroxylmethyl) aminomethane hydrochloride buffer (pH 7.6, 124 mM
NaCl, 25 mM CaCl2, 5 mM MnCl).
Data Analysis. All SPR Images were collected using the software
program XCAP v1.0 (EPIX Inc.). Further image analysis was performed
using the NIH Image v.1.61 software package. The line profile option
from this software was used to measure the signal (in pixel values)
from an SPR image, and this was subsequently converted to percent
reflectivity values. The error bars indicated in the figures show the
average percent error for all data points reported in the figure.
In section (F) of the Results and Discussion section, competitive
binding experiments are used to determine the KD values for lectin-
carbohydrate binding interactions. A multivalent Scatchard analysis is
used to characterize the interactions of multivalent acceptors with
multivalent ligands.24 This analysis can be used to construct an
expression for the equilibrium dissociation constant for the interaction
of a multivalent acceptor with a monovalent ligand under the
experimental conditions used in this study. In the following analysis,
the assumption is made that each acceptor site acts independently of
the other sites.
Experimental Methods
Materials. N-Succinimidyl S-acetylthiopropionate (SATP, Pierce),
2,2′-dipyridyl disulfide (DPDS, Aldrich), 11-mercaptoundecylamine
(MUAM, Dojindo Laboratories), concanavalin A (ConA, Sigma),
jacalin (Vector Laboratories), methyl R-D-mannopyranoside (Sigma),
methyl R-D-galactopyranoside (Sigma), and N-hydroxysuccinimidyl
ester of methoxypoly(ethylene glycol) propionic acid MW 2000
(Shearwater Polymers) were used as received. Experimental procedures
(20) Brockman, J. M.; Nelson, B. P.; Corn, R. M. Annu. ReV. Phys. Chem. 2000,
51, 41-63.
(25) Green, M. J.; Barner, B. J.; Corn, R. M. ReV. Sci. Instrum. 1991, 62, 1426-
(21) Frutos, A. G.; Corn, R. M. Anal. Chem. 1998, 70, 449A-455A.
(22) Brockman, J. M.; Frutos, A. G.; Corn, R. M. J. Am. Chem. Soc. 1999,
121, 8044-8051.
1430.
(26) Barner, B. J.; Green, M. J.; Saez, E. I.; Corn, R. M. Anal. Chem. 1991, 63,
55-60.
(23) Nelson, B. P.; Grimsrud, T. E.; Liles, M. R.; Goodman, R. M.; Corn, R.
M. Anal. Chem. 2001, 73, 1-7.
(24) Harris, S. J.; Jackson, C. M.; Winzor, D. J. Arch. Biochem. Biophys. 1995,
316, 20-23.
(27) Jordan, C. E.; Frutos, A. G.; Thiel, A. J.; Corn, R. M. Anal. Chem. 1997,
69, 4939-4947.
(28) Nelson, B. P.; Frutos, A. G.; Brockman, J. M.; Corn, R. M. Anal. Chem.
1999, 71, 3928-3934.
9
J. AM. CHEM. SOC. VOL. 125, NO. 20, 2003 6141

A R T I C L E S
Smith et al.
Figure 1. Schematic presentation of the surface modification process with SATP: (Step 1) MUAM surface; (Step 2) thiolation with SATP; (Step 3)
sulfhydryl deprotection; (Step 4) activated sulfhydryl surface with pyridyl groups; (Step 5) exchange reactions between the pyridyl disulfide surface and
sulfhydryl-containing carbohydrate derivative (R-SH) in solution, which produces the leaving group pyridine-2-thione. When the surface (Step 5) is exposed
to DTT, the disulfide linkage is cleaved. The immobilized molecule is released and the free sulfhydryl surface (Step 3) is regenerated.
The interaction between a monovalent ligand (L), and one acceptor
binding site (A) on a multivalent protein (P) can be represented as
and the resulting expression for the equilibrium dissociation constant
is
-1
L + A a LA
KA ) KD
[P][Lo]
KD )
.
(5)
[Po] - [P]
The expression for the equilibrium dissociation constant for this
interaction is
To determine KD, the SPR imaging data is used to measure [P] for
[L][A]
known values of [Po] and [Lo].
KD )
(1)
[LA]
Results and Discussion
where [LA] is the concentration of acceptor/ligand complexes, [A] is
the concentration of free acceptor sites, and [L] is the concentration of
free ligand. The concentration of total acceptor sites, [Ao], is equal to
the concentration of free acceptor sites plus the concentration of
acceptor/ligand complexes
A. Carbohydrate Surface Attachment Chemistry and
Surface Monolayer Characterization. To attach carbohydrate
derivatives to the surface, we chose to use a disulfide linkage
which has been previously shown to be effective for immobiliz-
ing thiol-modified oligonucleotides29 and peptides30 onto gold
surfaces. The reaction scheme that was used is depicted in Figure
1. A densely packed self-assembled monolayer of an amine-
terminated alkanethiol is first generated (MUAM, step 1), and
a series of successive reactions is used to modify this monolayer
via covalent bond formation. The MUAM surface is treated with
the heterobifunctional cross-linker S-acetylthiopropionate, which
contains an amine-reactive succinimide ester and a protected
thiol moiety. Reaction of the succinimide ester group with the
primary amine on MUAM results in the formation of an amide
bond to yield a monolayer with masked thiol groups (Step 2).
The thioester can be transformed to yield a free sulfhydryl
surface with an effective nucleophile, such as hydroxylamine
(Step 3). These reaction steps have been characterized with PM-
FTIRRAS.29
[Ao] ) [A] + [LA]
substituting this into eq 1
KD )
(2)
[A][L]
.
(3)
[Ao] - [A]
In surface competition experiments the total protein concentration [Po]
is constant, and the SPR response and adsorption isotherm are used to
determine the equivalent free protein concentration [P] as the amount
of ligand is increased. The concentration of the free acceptor sites in
solution, [A], is equal to q[P], where q is the number of binding sites
on a protein. Likewise, the total concentration of acceptor sites [Ao] is
equal to q[Po]. Substituting these values into eq 3 yields
q[P][L]
[P][L]
KD )
)
(4)
To implement the disulfide attachment strategy for carbohy-
drates, derivatives containing thiol groups needed to be gener-
q[Po] - q[P] [Po] - [P]
If the initial concentration of ligand is much greater than the initial
concentration of protein ([Lo] . [Po]), then the concentration of free
ligand is approximately the concentration of initial ligand ([L] ≈ [Lo]),
(29) Smith, E. A.; Wanat, M. J.; Cheng, Y.; Barreira, S. V. P.; Frutos, A. G.;
Corn, R. M. Langmuir 2001, 17, 2502-2507.
(30) Wegner, G. J.; Lee, H. J.; Corn, R. M. Anal. Chem. 2002, 74, 5161-5168.
9
6142 J. AM. CHEM. SOC. VOL. 125, NO. 20, 2003

SPR Imaging Studies of Protein-Carbohydrate Interactions
A R T I C L E S
Figure 2. Structure of the R-mannose (1) and R-galactose (2) derivatives
used in this study. Both were synthesized to contain an anomeric linker
with a free sulfhydryl group. The complete synthesis and characterization
details are provided in the Supporting Information.
Table 1. PM-FTIRRAS Band Assignment of Surface Modification
with Carbohydrate Probes
-1
surface
wavenumber (cm
)
assignment
Figure 3. Schematic presentation of the five-step process used to fabricate
carbohydrate arrays using PDMS microchannels to pattern the gold surface.
(Step 1) A thin gold film is immersed in an ethanolic solution of amine-
terminated alkanethiol to form a packed monolayer. (Step 2) PDMS
microchannels are placed on the gold surface. (Step 3) The carbohydrate
ligands are immobilized on the surface of the gold film using the attachment
chemistry presented in Figure 1. (Step 4) After immobilizing the carbohy-
drate, the PDMS is removed from the surface. The entire surface is then
reacted with a succinimide ester derivative of PEG to form a background
monolayer that resists the nonspecific adsorption of proteins from solution.
(Step 5) The interaction of the array with proteins in solution is monitored
with SPR imaging. Shown in the lower right corner is a typical SPR image
of a array fabricated using PDMS microchannels. Each yellow line
corresponds to one array element consisting of a unique carbohydrate or
mixture of carbohydrates.
mannose
or
galactose
derivative
3334 (broad)
2928
2855
1261
1191
OH stretching vibrations
CH2 asymmetrical stretch
CH2 symmetrical stretch
CH2 twisting mode
CO stretch of cyclic ether
CO stretch of dialkyl ether
1107
ated. We chose to append a thiol-containing linker at the
anomeric position. This site is a logical position for attachment
as it is used for naturally occurring glycoconjugates. The
mannose (1) and galactose (2) derivatives used in this study
were selected because they interact with different well-
characterized lectins (Figure 2). Specifically, R-linked mannose
derivatives bind to ConA, but galactose ligands do not.8 In
contrast, the lectin jacalin interacts preferentially with R-linked
galactose over mannose derivatives.31 By using derivatives 1
and 2, the ability of SPR imaging to distinguish specific
carbohydrate-protein interactions can be determined.
With the carbohydrate derivatives in hand, a two-step reaction
is used to attach thiol-containing carbohydrates onto the
sulfhydryl-terminated surface. The first step involves the forma-
tion of an immobilized mixed disulfide via reaction of the
surface sulfhydryl groups and 2,2′-dipyridyl disulfide in solution
(Step 4). The target compounds could then be appended through
a thiol-disulfide exchange reaction between the thiol-containing
carbohydrate and the surface disulfide (Step 5). The pyridine-
2-thione released in this process is a better leaving group than
the surface or the carbohydrate-containing thiolate; thus, the
desired disulfide is formed preferentially.32 The surface density
of immobilized ligands obtained with this attachment chemistry
is approximately 1013 molecules/cm2. This surface density is
similar to that obtained in the attachment of peptides to gold
surfaces using the disulfide attachment chemistry.30
from the sugar hydroxyl groups. Bands at 2928 and 2855 cm-1
were assigned to the CH2 asymmetric and symmetric stretching
modes of the linker, respectively. In the Mid-IR region, peaks
centered at 1261 cm-1 were assigned to the CH2 twisting modes.
The bands at 1191 and 1107 cm-1 have been assigned to the
CO stretch of the cyclic ether in the sugar ring and the CO
stretch of the dialkyl ether in the linker attachment, respectively.
B. Fabrication of Mannose and Galactose Arrays. Man-
nose and galactose arrays were fabricated using a combination
of the disulfide surface attachment chemistry and PDMS
microchannels to pattern the surface (Figure 3). To generate
the array, a MUAM monolayer was formed on the gold film
(Figure 3, Step 1). PDMS microchannels that had been
fabricated using a 3-D silicon mask were placed on the surface
(Figure 3, Step 2).33 These microchannels were composed of a
series of parallel channels, each with a width of 300 µm, that
had entrance and exit reservoirs at their ends for sample
introduction. Immobilization of the carbohydrates was carried
out in the channels by flowing 1 µL of solution through the
channels. Because each channel was independent, a different
carbohydrate or mixture of carbohydrates could be introduced
and immobilized within each channel (Figure 3, step 3). After
carbohydrate attachment, the PDMS microchannels were re-
moved from the modified surface. The number of species that
can be immobilized in this manner depends on the number of
channels fabricated. To date, up to 10 species have been
immobilized on a gold surface using this technique; however,
as many as 150 species can be immobilized on the same gold
PM-FTIRRAS spectra from the Mid-IR and CH stretching
regions (2500-3500 cm-1) were acquired to verify the attach-
ment of these carbohydrate derivatives to the surface. Table 1
lists the FTIR peak assignments in the Mid-IR and CH stretching
regions for the spectra obtained upon attaching compounds 1
and 2 to the surface using the disulfide linkage. The broad peak
at 3334 cm-1 has been assigned to the OH stretching vibrations
(31) Bourne, Y.; Astoul, C. H.; Zamboni, V.; Peumans, W. J.; Menu-Bouaouiche,
L.; Van Damme, E. J. M.; Barre, A.; Rouge, P. Biochem. J. 2002, 364,
173-180.
(33) Lee, H. J.; Goodrich, T. T.; Corn, R. M. Anal. Chem. 2001, 73, 5525-
(32) Grasetti, D. R.; Murray, J. F. J. Anal. Biochem. 1967, 21, 427.
5531.
9
J. AM. CHEM. SOC. VOL. 125, NO. 20, 2003 6143

A R T I C L E S
Smith et al.
The carbohydrate arrays fabricated with this attachment
chemistry were robust, and could be used in approximately 20
binding cycles before degradation of the surface was observed.
The arrays were regenerated by treatment with a high pH buffer
solution (pH 11.5) and/or high concentration salt solutions to
remove the bound lectin. The major factor limiting the use of
these arrays in additional binding experiments is the slow
accumulation of nonspecifically adsorbed protein onto the
surface.The image shown on the right in Figure 4 and corre-
sponding line profile (solid line) were obtained after removing
the ConA from the surface and exposing the array to a 100 nM
solution of the lectin jacalin. In contrast to ConA, jacalin was
found to bind selectively to the galactose derivative. The
selectivity of ConA and jacalin toward the immobilized
carbohydrates demonstrates that this immobilization method
affords surfaces suitable for assessing the protein binding
specificity for carbohydrates.
There are three factors that can influence the interaction
between the adsorbing protein and the carbohydrate surface:
protein-protein interactions between adsorbing protein mol-
ecules, the carbohydrate surface site density, and the number
of interactions occurring between the adsorbing protein and the
carbohydrate surface. The protein-protein interactions were
studied using adsorption isotherms (section C). To study the
effect of the surface site density on protein adsorption, the
composition of the active carbohydrate immobilized on the
surface was varied (section D). Finally, the effect of multivalent
protein binding to the surface was studied by constructing
adsorption isotherms for surfaces with varying carbohydrate
compositions (section E).
C. Adsorption Coefficients for the Binding of Jacalin to
Galactose Surfaces and ConA to Mannose Surfaces. To study
the interaction strength between the immobilized carbohydrates
and ConA or jacalin, adsorption isotherms for these lectins
interacting with carbohydrate arrays were constructed by
monitoring the protein surface coverage as a function of protein
concentration. Thermodynamic data suggest that the binding
affinity of jacalin to methyl-R-D-galactopyrannoside (R-MeGal,
KD ) 25 µM)37 is higher than that for ConA binding to methyl-
R-D-mannopyrannoside (R-MeMan, KD ) 130 µM).38 Conse-
quently, we sought to determine if differences in protein
adsorption could be detected.
Figure 4. Line profiles taken from SPR difference images of a two-
component carbohydrate array with two (left) channels containing compound
1 and two (right) channels containing compound 2. The SPR image on the
left and dotted line profile below correspond to the introduction of the lectin
ConA to the surface and show the binding of ConA to the mannose array
elements. The SPR image on the right and solid line profile correspond to
the introduction of the lectin jacalin to the surface and show the binding of
jacalin to the galactose array elements.
film (1.8 × 1.8 cm) by reducing the channel width and the space
between the channels.33
To form an array background that is resistant to protein
adsorption, the entire surface was exposed to a solution of the
succinimide ester of poly(ethylene glycol) in phosphate buffer
(pH 8.0). The succinimide ester reacts with unmodified amine
groups to generate a surface covered in poly(ethylene glycol)
groups (Figure 3, step 4). Such surfaces have been shown to
resist nonspecific protein adsorption.34-36 The resulting array
was then placed into a flow cell for SPR imaging experiments.
A typical SPR image of a four-component carbohydrate array
is shown in Figure 3 (Figure 3, step 5).
To determine whether the carbohydrates immobilized using
the disulfide attachment chemistry are accessible to proteins in
solution, a two-component carbohydrate array was fabricated.
The resulting array, which contains two channels of the mannose
derivative (1), and two channels of the galactose derivative (2)
was exposed to the lectins ConA and jacalin. As mentioned
previously, the lectin ConA exhibits selectivity for mannose over
galactose residues;8 jacalin binds to both galactose and mannose,
but it binds to mannose residues with several orders of
magnitude lower affinity than to galactose residues.31 Shown
in Figure 4 are two SPR difference images, obtained by
subtracting the images taken before and after the protein
adsorption events, and the corresponding line profiles taken from
these SPR images. The image on the left and corresponding
line profile (dotted line) were obtained upon exposing the array
to a 200 nM solution of ConA. The data indicate that the protein
binds selectively to the mannose array elements.
It has been previously shown that SPR imaging can be used
to quantitate the amount of material adsorbing to a gold surface;
thereby providing a measure of protein surface coverage.23 The
adsorption isotherms for the binding of (Figure 5, circles) ConA
to a mannose surface, and (Figure 5, squares) the binding of
jacalin to a galactose surface were obtained by plotting the
relative protein surface coverage at equilibrium as a function
of the protein concentration. The solid lines in Figure 5 were
obtained by fitting the data to a Frumkin isotherm, which has
the form
θ
-1
[P] ) KADS
e-2aθ
(6)
1 - θ
where [P] is the concentration of protein in solution, KADS is
(34) Silin, V.; Weetall, H.; Vanderah, D. J. J. Coll. Inter. Sci. 1997, 185, 94-
103.
(35) Ostuni, E.; Chapman, R. G.; Holmlin, R. E.; Takayama, S.; Whitesides,
G. M. Langmuir 2001, 17, 5605-5620.
(37) Mahanta, S. K.; Sastry, M. V.; Surolia, A. Biochem. J. 1990, 265, 831-
840.
(38) Weatherman, R. V.; Mortell, K. H.; Chervenak, M.; Kiessling, L. L.; Toone,
E. J. Biochem. 1996, 35, 3619-3624.
(36) Harder, P.; Grunze, M.; Dahint, R.; Whiteside, G. M.; Laibinis, P. E. J.
Phys. Chem. B 1998, 102, 426-436.
9
6144 J. AM. CHEM. SOC. VOL. 125, NO. 20, 2003

SPR Imaging Studies of Protein-Carbohydrate Interactions
A R T I C L E S
Figure 5. Plots of the relative protein surface coverage (fraction of occupied
surface sites, θ) as determined by SPR imaging as a function of the protein
concentration in the solution. The data shown correspond to (squares) the
binding of jacalin to a galactose surface and (circles) the binding of ConA
to a mannose surface. The data have been fit to Frumkin isotherms (solid
lines).
Figure 6. Plots showing the normalized SPR signal obtained for (square)
ConA and (triangle) jacalin binding to surfaces with the indicated composi-
tions of mannose and galactose derivatives. See text for further discussion.
the adsorption coefficient, θ is the protein surface coverage
(given by Γ/ΓMAX), and a is the Frumkin interaction parameter.39
The Frumkin interaction parameter in eq 6 indicates whether
the adsorbing molecules exhibit attractive or repulsive (lateral)
interactions. When a is equal to zero in eq 6, the Langmuir
adsorption isotherm is obtained. Positive Frumkin interaction
parameters were obtained from the fits of the lectin isotherms
shown in Figure 5, and this indicates attractive forces are present
between the adsorbing protein molecules. The adsorption
coefficient provides information about the strength of the
interaction between the adsorbing species and the species on
the surface. The adsorption coefficient for jacalin binding to
the galactose-substituted surface was found to be 2.2 ( 0.8 ×
107 M-1, whereas that for the interaction of ConA and the
mannose-derivatized surface was found to be 5.6 ( 1.7 × 106
M-1. The adsorption coefficients indicate that the interaction
between jacalin and the surface immobilized galactose is
stronger than the interaction between ConA and the surface
immobilized mannose, which is consistent with the aforemen-
tioned thermodynamic data.
D. Mixed Galactose and Mannose Monolayers. Both ConA
and jacalin are tetrameric proteins, with 4 carbohydrate-binding
sites. Thus, we envisioned that these proteins might engage in
multivalent interactions with the appropriate surfaces. To study
multivalent binding between the lectin and the surface, SPR
imaging was used to monitor the binding of ConA and jacalin
to surfaces that presented varying compositions of immobilized
carbohydrates. The composition of the immobilized carbohydrate
surfaces can be controlled using the disulfide attachment scheme
by varying the concentration of the carbohydrate derivatives
that are conjugated to the surface (Figure 3, step 3). The active
carbohydrate derivative (e.g., compound 1 for ConA) can be
diluted with a nonactive species (e.g., compound 2 for ConA).
We choose to use carbohydrates for both the “active” and
“inactive” epitopes as the sterics of the surface should not be
altered. Thus, adsorption data can be more readily compared.40
Shown in Figure 6 are the normalized SPR signals obtained
for the binding of (squares) ConA and (triangles) jacalin to
mixed carbohydrate monolayers, composed of mixtures of
mannose and galactose ligands. The amount of mannose or
galactose immobilized on the surface varied in these studies,
however the total amount of carbohydrate remained constant.
For example, the 0% mannose surface contained 100% galac-
tose, and the 50% mannose surface contained 50% galactose.
The data provide insight into the interactions of the lectins with
the carbohydrate-modified surfaces.
The binding of either ConA or jacalin to the surface could
be detected by SPR imaging, even for very low surface
compositions of active carbohydrate. For ConA, protein binding
to surfaces composed of 0.2% mannose was detected. In
addition, there is a linear relationship between the amount of
protein adsorbed to the surface and the mannose composition
for the binding of ConA to surfaces presenting low mannose
compositions (0 to 1% mannose). For surfaces with composi-
tions of mannose greater than 1%, the relationship is no longer
linear. The binding of a 150 nM solution of jacalin to 0%
galactose surfaces showed that a small but significant amount
of protein adsorbed to the surface; this binding was not observed
at low (below 100 nM) jacalin concentrations. This result is
consistent with reports that jacalin has a low affinity for
mannose, albeit an affinity that is several orders of magnitude
lower than its affinity for galactose. In contrast, no significant
ConA binding to the 0% mannose surface occurs even at a ConA
concentration of 300 nM.
The data also indicate that there is a plateau in the amount
of ConA or jacalin adsorbing to the surface as the amount of
active carbohydrate on the surface increases (Figure 6). These
experiments were performed at protein concentrations corre-
sponding to less than a full monolayer of protein (300 nM ConA
and 150 nM jacalin) for the 100% mannose and galactose
surfaces. Therefore, the plateau in protein adsorption cannot be
explained by steric constraints between adsorbed protein
molecules that would limit the amount of protein that adsorbs
to the surface. In addition, it has been previously shown by
SPR imaging that the binding of antibody from solution onto
mixed peptide monolayers containing a peptide sequence
recognized by the antibody results in a linear relationship
between the amount of antibody binding to the surface and the
amount of active peptide immobilized on the surface.30
The plateau in protein binding seen in Figure 6 indicates that
there is a density of immobilized carbohydrate ligands beyond
which the addition of more binding epitopes does not increase
the adsorption of protein to the surface. This result suggests
(39) Adamson, A. W.; Gast, A. P. Physical Chemistry of Surfaces; John Wiley
& Sons Inc.: New York, 1997.
(40) Cairo, C. W.; Gestwicki, J. E.; Kanai, M.; Kiessling, L. L. J. Am. Chem.
2002, 124, 1615-1619.
9
J. AM. CHEM. SOC. VOL. 125, NO. 20, 2003 6145

A R T I C L E S
Smith et al.
Figure 7. Schematic model of the data shown in Figure 6. For a low level
of ligand substitution on the surface, the spacing between the carbohydrate
binding sites on the protein prohibits the protein from binding to more than
one immobilized carbohydrate epitope. At an intermediate ligand density,
the spacing of the carbohydrates on the surface allows the protein to interact
with multiple carbohydrates on the surface. At a high carbohydrate coverage,
immobilization additional carbohydrates does not increase the amount of
protein bound to the surface.
Figure 8. Isotherms for jacalin binding to (triangle) 10%; (square) 50%;
and (circle) 100% galactose-modified surfaces. Each set of data has been
normalized to the maximum signal obtained for that surface, and was fit
using the Frumkin isotherm equation. Shown in Table 2 are the Frumkin
interaction parameters and adsorption coefficients for jacalin interacting
with these surfaces.
Table 2. Langmuir Adsorption Coefficients (KADS), and Frumkin
Interaction Parameters
that at high ligand densities not all of the carbohydrate residues
at high mannose- (ConA) or galactose-derivatized (jacalin)
surfaces are active. This observation can be explained by the
distance between the carbohydrate binding domains on the
lectins and the density of the immobilized carbohydrates on the
surface. For example, ConA is capable of forming two attach-
ment points to the surface, and the distance between these points
on the protein is approximately 6.5 nm.41 On the basis of the
surface density of ligands immobilized using the disulfide
attachment chemistry for 10%, 50%, and 100% mannose
surfaces, for example, the approximate distance between man-
nose residues is 11, 5, and 4 nm, respectively. For a 10%
mannose-substituted surface, the average distance between
immobilized mannose residues is too great for adsorbed ConA
to be interacting with more than one mannose residue (Figure
7, left). Near 50% mannose substitution, the mannose residues
are close enough on the surface such that on average an adsorbed
ConA can bind to two mannose residues (Figure 7, center).
When the level of mannose-substitution is 100%, there are more
than 2 carbohydrate derivatives available on the surface within
the carbohydrate binding region of each protein; however, each
protein can only interact with 2 mannose residues at one time
(Figure 7, right). No further protein can adsorb to the surface
even though some mannose residues remain unoccupied due to
the limitations on the carbohydrate binding site density.
E. Adsorption Isotherms for Lectin binding to 10%, 50%,
and 100% Carbohydrate Surfaces. To determine if the
interaction strength between the adsorbed protein and the surface
changes as the amount of carbohydrate ligand on the surface
increases, the adsorption isotherms for jacalin and ConA were
constructed for surfaces composed of 10%, 50%, and 100%
carbohydrate ligand. If the lectins engage in multivalent binding
with the surface at high levels of ligand substitution, then the
interaction strength between the lectin and the surface should
increase as the percentage of ligand increases. The adsorption
isotherms for jacalin binding to (triangles) 10%, (squares) 50%,
and (circles) 100% galactose surfaces (Figure 8) were fit to
Frumkin isotherms (eq 6), and the Frumkin interaction param-
eters and adsorption coefficients obtained from these data are
protein
surface composition
KADS
a
jacalin
100% galactose
50% galactose
10% galactose
2.2 ( 0.8 × 107 M-1
1.6 × 107 M-1
0.5 ( 0.3
0.4
0.4
1.2 × 107 M-1
ConA
100% mannose
50% mannose
10% mannose
5.6 ( 1.7 × 106 M-1
0.55
0.55
0
5.6 × 106 M-1
5.0 × 106 M-1
listed in Table 2. The adsorption coefficients indicate that the
interaction between jacalin and the surface becomes stronger
as the amount of immobilized galactose increases on the surface.
The adsorption isotherms for ConA binding to 10%, 50%,
and 100% mannose surfaces show similar trends to the jacalin
adsorption isotherms (Table 2). The interaction strength in-
creases between ConA and the mannose-substituted surface upon
increasing the composition of mannose from 10% to 50%. The
adsorption coefficient remains constant, within experimental
error, as the mannose composition is further increased to 100%.
The increase in KADS for surfaces presenting high densities of
active carbohydrate imply that at high active carbohydrate
densities, the lectins are interacting with the surface through
multivalent interactions, and that the extent of multivalent
binding increases as the surface density of active carbohydrate
increases.
F. Solution Equilibrium Dissociation Constants of Car-
bohydrate-Lectin Interactions Measured with SPR Imaging.
It is possible to determine the solution equilibrium dissociation
constants (KD) for carbohydrate lectin interactions using the SPR
imaging technique to probe the concentration of unbound protein
in the presence of carbohydrate in solution. This analysis allows
for the direct comparison of affinity measurements obtained with
SPR imaging and solution based affinity measurements.
In a competitive binding experiment, carbohydrates in solution
compete with the immobilized carbohydrate ligands for binding
sites on the lectin, establishing a coupled equilibrium between
the binding of the protein to the surface immobilized species
and to the species present in solution. This is schematically
represented in the inset to Figure 9. The total concentration of
(41) Bittiger, H.; Schnebli, H. P. ConcanaValin A as a Tool; John Wiley and
Sons: London, New York, 1976.
9
6146 J. AM. CHEM. SOC. VOL. 125, NO. 20, 2003

SPR Imaging Studies of Protein-Carbohydrate Interactions
A R T I C L E S
Figure 9. Plot of the relative surface coverage of jacalin, measured with
SPR imaging, as a function of the concentration of R-MeGal in solution.
The R-MeGal in solution competes with the surface immobilized galactose
for the jacalin binding sites through a coupled equilibrium shown in the
figure inset. The figure inset has been simplified to show jacalin binding
to one galactose molecule from solution; however, jacalin has four galactose
binding sites. The data are fit as discussed in the text, using eqs 5 and 6,
to determine the solution dissociation constant KD.
Figure 10. Plot of the relative surface coverage of ConA, measured with
SPR imaging, as a function of the concentration of R-MeMan in solution.
The R-MeMan in solution competes with the surface immobilized mannose
through a coupled equilibrium shown in the figure inset. The figure inset
has been simplified to show ConA binding to one mannose molecule from
solution; however, ConA has four mannose binding sites. The data are fit
as discussed in the text, using eqs 5 and 6.
KADS-1 is the protein concentration at which half of the surface-
binding sites are occupied with protein, and KD is the ligand
concentration at which half the protein-binding sites are occupied
with ligand in solution. In the Experimental Methods section,
the multivalent Scatchard analysis was used to derive an
equation for the equilibrium dissociation constant (eq 5),
whereas an adsorption isotherm was used in Section C to
characterize protein interactions with the surface. A comparison
of the equations that result from the Scatchard analysis and the
Langmuir adsorption isotherm illustrate the differences between
protein ([Po]) is held constant, while the concentration of
carbohydrate in solution ([Lo]) is increased. An associated
decrease in the amount of protein binding to the surface is
observed, which is monitored by SPR imaging (Figure 9). Using
the SPR imaging signal, the unbound protein concentration ([P])
can be obtained via the adsorption isotherm (eq 6). Once the
concentration of P has been measured, it is possible to determine
KD using eq 5, which is derived from the multivalent Scatchard
formula.24 This analysis makes three assumptions: (i) the
amount of protein binding to the sensor surface is negligible
compared to the total amount of protein in the system, (ii) the
initial concentration of ligand is much greater than the initial
concentration of protein so that the concentration of unbound
ligand is approximately equal to the total concentration of ligand
([Lo] ≈ [L]), and (iii) each monomer in the protein binds to the
ligand independently of the other monomers. A competition
binding experiment was performed with jacalin binding to a
1% galactose surface using R-MeGal in solution as the competi-
tor ligand (Figure 9). When the data is analyzed according to
eqs 5 and 6, a KD value of 16 ( 5 µM for jacalin and R-MeGal
was obtained. This value agrees well with the solution dissocia-
tion constant for jacalin binding to R-MeGal, which is 25 µM.37
The results for an inhibition binding experiment performed for
the binding of ConA to a 10% mannose surface with increasing
concentration of R-MeMan in solution are shown in Figure 10.
An equilibrium dissociation constant of 200 ( 50 µM was
obtained from this experiment. This value correlates well with
the dissociation constant of 130 µM for R-MeMan obtained by
microcalorimetry for the interaction of ConA with mannose.38
For both proteins, the lowest surface composition of active
carbohydrate that resulted in a detectable signal upon protein
binding inhibition was chosen for the competition binding
experiments.
the parameters KD and KADS
.
For the interaction between a monovalent ligand (L) and a
multivalent protein (P) with q binding sites, the total concentra-
tion of bound ligand divided by the total concentration of protein
is defined by Winzor et al.24 as the parameter rf
[Lo] - [L]
rf )
(7)
[Po]
where [Lo] is the total concentration of ligand, [L] is the
concentration of free ligand, and [Po] is the total concentration
of protein in the system. The assumptions for the analysis
presented here are the same as those listed previously. Equation
7 can be rewritten in terms of the total [Ao] and free [A] acceptor
site concentrations considering that (i) [Ao] ) q[Po]; (ii) [Ao]
) [A] + [LA]; and (iii) [Lo] ) [L] + [LA]
[Ao] - [A]
[Ao]/q
[A]
rf )
) q 1 -
(8)
[
]
[Ao]
Rearrangement of eq 3 to obtain an equation for [A]/[Ao]
[A]
1
)
(9)
1 + KD-1[L]
[Ao]
G. KD versus KADS: A Comparison of the Multivalent
Scatchard Analysis and the Langmuir Adsorption Isotherm.
In the previous section, we employed competitive binding
measurements to calculate KD, the dissociation constant for
ligand-protein interactions in solution. This equilibrium con-
stant is different than the KADS that was measured in section C
for protein adsorption to the surface. Specifically, the parameter
and substitution of eq 9 into eq 8 yields
KD-1[L]
rf ) q
(10)
-1
[
]
1 + KD [L]
The parameter rf is the average number of occupied ligand
9
J. AM. CHEM. SOC. VOL. 125, NO. 20, 2003 6147

A R T I C L E S
Smith et al.
binding sites on a protein molecule in solution. Equation 10
has the same functional form as the Langmuir adsorption
isotherm (eq 6 with a ) 0)
sociation constants for the lectins ConA and jacalin. These
studies also highlight the utility of changing the ligand density
on gold films as a tool to study multivalent interactions. We
anticipate that the methods for fabricating carbohydrate arrays
via the disulfide attachment chemistry combined with SPR
imaging will facilitate the study of protein-carbohydrate
interactions. We envision these methods will be useful for
characterizing the sugar binding specificities of lectins and
antibodies, identifying glycomimetics, and in carrying out high-
throughput assays for the discovery of compounds that enhance
or disrupt lectin-carbohydrate interactions.
K
ADS[P]
θ )
(11)
1 + KADS[P]
where θ is the fraction of surface sites occupied with protein
molecules, and KADS is the adsorption coefficient. Equations
10 and 11 quantitatively show the difference between KD and
KADS-1. A comparison of the KD (16 µM and 200 µM for jacalin
-1
and ConA, respectively) and the KADS (45 nM and 180 nM
Acknowledgment. This research is funded by the NSF (Grant
CHE-0133151 to R.M.C.), the Defense Advanced Research
Projects Agency (DARPA) and the Air Force Research Labora-
tory, Air Force Material Command, USAF, (F30602-01-2-0555
to R.M.C.), the NIH (Grant GM59622 to R.M.C.), and (Grant
GM49975 to L.L.K.). We thank the W.M. Keck Foundation
for support.
for jacalin and ConA, respectively) values shows that although
the adsorption coefficients can be used to rank the carbohydrate-
lectin interaction strengths in solution, they cannot be used to
determine directly the solution KD values.
Conclusions
We have shown that SPR imaging can be used to study
characteristically weak protein-carbohydrate interactions, and
that this method provides the means to assess the carbohydrate
binding selectivity of lectins. SPR imaging can be used to
quantitate the strength of lectin-carbohydrate interactions by
determining the adsorption coefficients and the solution dis-
Supporting Information Available: Complete synthesis and
characterization details for compounds 1 and 2. This material
is available free of charge via the Internet at http://pubs.acs.org.
JA034165U
9
6148 J. AM. CHEM. SOC. VOL. 125, NO. 20, 2003