Probing low affinity and multivalent interactions with surface plasmon resonance: Ligands for concanavalin A

Article abstract of DOI:10.1021/ja9818506

The affinities of the carbohydrate-binding protein concanavalin A (Con A) for mono- and multivalent ligands were measured by surface plasmon resonance (SPR) detection. Assessing protein-carbohydrate affinities is typically difficult due to weak affinities observed and the complications that arise from the importance of multivalency in these interactions. We describe a convenient method to rapidly evaluate the inhibitory constants for a panel of different ligands, both monovalent and multivalent, for low- affinity receptors, such as the carbohydrate-binding protein Con A. A nonnatural, mannose-substituted glycolipid was synthesized, and self- assembled monolayers of varying carbohydrate density were generated. The synthetic surfaces bind Con A. Competition experiments that employed monovalent ligands in solution yielded K(i) values similar to equilibrium binding constants obtained in titration microcalorimetry experiments. In addition, this assay could be used to examine various polymeric ligands of defined lengths, generated by ring-opening metathesis polymerization (ROMP). This study demonstrates the utility of this method for rapidly screening ligands that engage in low affinity interactions with their target receptors. Our results emphasize that those molecules that can simultaneously occupy two or more saccharide binding sites within a lectin oligomer are effective inhibitors of protein-carbohydrate interactions.

Full text of DOI:10.1021/ja9818506

VOLUME 120, NUMBER 41
OCTOBER 21, 1998
©
Copyright 1998 by the
American Chemical Society
Probing Low Affinity and Multivalent Interactions with Surface
Plasmon Resonance: Ligands for Concanavalin A
David A. Mann, Motomu Kanai, Dustin J. Maly, and Laura L. Kiessling*
Contribution from the Departments of Chemistry and Biochemistry, UniVersity of Wisconsin,
Madison, Wisconsin 53706
ReceiVed May 27, 1998
Abstract: The affinities of the carbohydrate-binding protein concanavalin A (Con A) for mono- and multivalent
ligands were measured by surface plasmon resonance (SPR) detection. Assessing protein-carbohydrate affinities
is typically difficult due to weak affinities observed and the complications that arise from the importance of
multivalency in these interactions. We describe a convenient method to rapidly evaluate the inhibitory constants
for a panel of different ligands, both monovalent and multivalent, for low-affinity receptors, such as the
carbohydrate-binding protein Con A. A nonnatural, mannose-substituted glycolipid was synthesized, and self-
assembled monolayers of varying carbohydrate density were generated. The synthetic surfaces bind Con A.
Competition experiments that employed monovalent ligands in solution yielded Ki values similar to equilibrium
binding constants obtained in titration microcalorimetry experiments. In addition, this assay could be used to
examine various polymeric ligands of defined lengths, generated by ring-opening metathesis polymerization
(ROMP). This study demonstrates the utility of this method for rapidly screening ligands that engage in low
affinity interactions with their target receptors. Our results emphasize that those molecules that can
simultaneously occupy two or more saccharide binding sites within a lectin oligomer are effective inhibitors
of protein-carbohydrate interactions.
Introduction
as substrates for protein binding and the weak affinity observed
in monovalent interactions have led to speculation that the
clustering of binding events into multivalent arrays leads to a
greater affinity and specificity than predicted from the sum of
Numerous cellular recognition processes depend on protein-
carbohydrate interactions. These lectin-ligand attachments are
critical in fertilization, cell signaling, pathogen identification,
and the inflammatory response.1 Lectin binding of carbohy-
drate ligands is enigmatic because the attractive forces at work
5
-9
the constitutive one-on-one interactions.
Treating human
,2
diseases that involve protein-carbohydrate interactions is
challenging since targeting a specific recognition event requires
3
-4
-1
are noncovalent and relatively weak (i.e., Ka ≈ 10
M ),
(6) (a) Mortell, K. H.; Weatherman, R. V.; Kiessling, L. L. J. Am. Chem.
yet the strength and specificity required for proper cellular
Soc. 1996, 118, 2297-8. (b) Mortell, K. H.; Gingras, M.; Kiessling, L. L.
3-5
targeting is great.
The functional similarity of carbohydrates
J. Am. Chem. Soc. 1994, 116, 12053-4.
(7) Sauter, N. K.; Bednarski, M. D.; Wurzburg, B. A.; Hanson, J. E.;
(
4.
1) Drickamer, K.; Taylor, M. E. Annu. ReV. Cell Biol. 1993, 9, 237-
Whitesides, G. M.; Skehel, J. J.; Wiley: D. C. Biochemistry 1989, 28, 8388-
96.
6
(
(
(
(
2) Varki, A. Glycobiology 1993, 3, 97-130.
(8) Crocker, P. R.; Feizi, T. Curr. Opin. Struct. Biol. 1996, 6, 679-91.
(9) 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, 1520-2.
3) Kiessling, L. L.; Pohl, N. L. Chem. Biol. 1996, 3, 71-7.
4) Roy, R. Curr. Opin. Struct. Biol. 1996, 6, 692-702.
5) Lee, Y.; Lee, R. Acc. Chem. Res. 1995, 28, 321-7.
1
0.1021/ja9818506 CCC: $15.00 © 1998 American Chemical Society
Published on Web 10/06/1998

10576 J. Am. Chem. Soc., Vol. 120, No. 41, 1998
Mann et al.
3
6
-1 24
an intimate knowledge of factors leading to lectin-ligand
specificity. Understanding these key elements will facilitate the
development of new therapeutic strategies.
(BIAcore 2000) is reported to be 10 -10 M . Association
constants in this range approximate the low affinities observed
in many protein-carbohydrate complexes, as is seen for the
jack bean lectin concanavalin A (Con A) binding to mono- and
The recent automation of surface plasmon resonance (SPR)
technology, which can be used to measure affinity rate constants,
allows the convenient application of this valuable technique for
assessing the rates of interaction. This method has been used
6
,23,25
multivalent glycopyranosides.
Complexation of the low
molecular weight (200 Daltons) monosaccharide ligands to
immobilized protein would be difficult to detect by SPR;
consequently, lectin binding to surface-bound carbohydrate was
monitored.
Many investigations employing SPR have monitored the
binding of an immobilized ligand to a target receptor in solution.
One thorough study in which several receptor-ligand interac-
tions were analyzed indicated that the off-rates for receptor
dissociation from a surface are often unrelated to the dissociation
rate in solution. Thus, the precedents indicate that apparent
affinities determined for ligand-modified surfaces do not cor-
relate with solution binding constants. Moreover, the features
of the synthetic surfaces can influence the observed binding
interactions, complicating attempts to determine the binding of
ligand in solution by comparing surface binding.
Several recent studies identify distinct advantages of competi-
tion assays in which the ability of a substrate to inhibit the
interactions of a soluble receptor with an immobilized ligand,
10-12
to analyze a multitude of ligand-ligate complexes.
Using
an optical biosensor, it is easy to determine the apparent rates
of association and dissociation at a surface by monitoring free
ligand binding to an immobilized binding partner. High affinity
mono- and multivalent protein-carbohydrate interactions have
13-19
been studied previously using SPR.
These assays, however,
do not allow for quantitative, rapid screening of multiple low-
affinity ligands. In each case, a new surface must be created
for presentation and subsequent assessment of each ligand to
be tested. Despite its potential, SPR has not been applied to
the evaluation of multivalent inhibitors.
We have developed an SPR competition binding assay to
garner quantitative binding data on monovalent and multivalent
lectin-ligand complexes. Our interest is in understanding how
weak, low-affinity interactions are used physiologically to
achieve enhanced affinity and specificity. To this end, synthetic,
2
6
2
6-29
multivalent glycoprotein mimics, termed neoglycopolymers,
is explored.
In such assays, a ligand competes for a
were devised to probe this issue.6
,20-23
In search of a rapid,
receptor in solution, thereby minimizing differences associated
with surface composition. In addition to this advantage, only
one surface is needed to measure binding to a variety of
inhibitors. Consequently, competition assays were used to
determine the efficacy of several inhibitors, using protocols
similar to those reported by Karlsson and Morelock et al. 2
A mannose-derivatized glycolipid, 1, was synthesized (Figures
sensitive, and reproducible assay requiring modest amounts of
lectin and ligand, we evaluated SPR detection for these purposes.
The sensitivity limit for Ka determination using a com-
mercially available surface plasmon resonance instrument
6,29
(
10) (a) van der Merwe, P. A.; Barclay, A. N. Curr. Opin. Immunol.
996, 8, 257-61. (b) Szabo, A.; Stolz, L.; Granzow, R. Curr. Opin. Struct.
Biol. 1995, 5, 699-705. (c) Myszka, D. G. Curr. Opin. Biotechnol. 1997,
, 50-7.
11) For critical analyses of methods for collecting and analyzing SPR
1
1
and 2) and noncovalently bound to an optical sensor chip
8
(
surface through lipid bilayer formation. By combining the
synthetic glycolipid with phosphatidylcholine (POPC) in various
molar ratios, it was possible to control the density of saccharide
data, see: (a) Schuck, P. Annu. ReV. Biophys. Biomol. Struct. 1997, 26,
5
4
41-66. (b) Schuck, P.; Minton, A. P. Trends Biochem. Sci. 1996, 21,
58-60.
1
3,30
ligands presented.
Solution competition studies using the
(
12) For a description of the SPR instrument design and principles of
its operation, see: Garland, P. B. Q. ReV. Biophys. 1996, 29, 91-117.
13) Kuziemko, G. M.; Stroh, M.; Stevens, R. C. Biochemistry 1996,
5, 6375-84.
14) Imata, H.; Kubota, K.; Hattori, K.; Masaaki, A.; Jindoh, C. Bioorg.
Med. Chem. Lett. 1997, 7, 109-12.
15) (a) MacKenzie, C. R.; Hirama, T.; Deng, S.-J.; Bundle, D. R.;
glycolipid surface proved an excellent means for rapid genera-
tion of inhibition data, both with small, low-affinity monovalent
and larger, high-affinity multivalent Con A ligands.
(
3
(
(
Results
Narang, S. A.; Young, N. M. J. Biol. Chem. 1996, 271, 1527-33. (b) Deng,
S.-J.; MacKenzie, C. R.; Hirama, T.; Brosseau, R.; Lowary, T. L.; Young,
N. M.; Bundle, D. R.; Narang, S. A. Proc. Natl. Acad. Sci. U.S.A. 1995,
Initial attempts to develop an assay involved the use of a
gold surface coated with dextran, a polymer composed of
9
2, 4992-6. (c) Rheinnecker, M.; Hardt, C.; Ilag, L. L.; Kufer, P.; Gruber,
R.; Hoess, A.; Lupas, A.; Rottenberger, C.; Pl u¨ ckthun, A. J. Immunol. 1996,
57, 2989-97.
16) (a) Shinohara, Y.; Kim, F.; Shimizu, M.; Goto, M.; Tosu, M.;
(23) For alternative approaches to the synthesis of multidentate ligands
for concanavalin A, see ref 25b and (a) Roy, R.; Page, D.; Perez, S. F.;
Bencomo, V. V. Glycoconjugate J. 1998, 15, 251-63 and references therein;
(b) Kurita, K.; Kobayashi, M.; Munakata, T.; Ishii, S.; Nishimura, S. H.
Chem. Lett. 1994, 2063-6. (c) Lehmann, J.; Weitzel, U. P. Carbohydr. Res.
1996, 294, 65-94. (d) Matsuda, K.; Inazu, T.; Haneda, K.; Mizuno, M.;
Yamanoi, T.; Hattori, K.; Yamamoto, K.; Kumagai, H. Bioorg. Med. Chem.
Lett. 1997, 7, 2353-6.
(24) Hall, D. R.; Cann, J. R.; Winzor, D. J. Anal. Biochem. 1996, 235,
175-84.
(25) (a) Weatherman, R. V.; Kiessling, L. L. J. Org. Chem. 1996, 61,
534-8. (b) Weatherman, R. W.; Mortell, K. H.; Chervenak, M.; Kiessling,
L. L.; Toone, E. J. Biochemistry 1996, 35, 3619-24 and ref 36.
(26) Nieba, L.; Krebber, A.; Pl u¨ ckthun, A. Anal. Biochem. 1996, 234,
155-65.
(27) (a) Karlsson, R. Anal. Biochem. 1994, 221, 142-51. (b) Karlsson,
R.; Jendeberg, L.; Nilsson, B.; Nilsson, J.; Nygren, P. J. Immunol. Methods
1995, 183, 43-9.
(28) (a) Ward, L. D.; Howlett, G. J.; Hammacher, A.; Weinstock, J.;
Yasukawa, K.; Simpson, R. J.; Winzor, D. J. Biochemistry 1995, 34, 2901-
7. (b) O’Shannessy, D.; Winzor, D. Anal. Biochem. 1996, 236, 275-83.
(29) Morelock, M. M.; Ingraham, R. H.; Betageri, R.; Jakes, S. J. Med.
Chem. 1995, 38, 1309-18.
1
(
Hasegawa, Y. Eur. J. Biochem. 1994, 223, 189-94. (b) Shinohara, Y.; Sota,
H.; Kim, F.; Shimizu, M.; Gotoh, M.; Tosu, M.; Hasegawa, Y. J.
Biochemistry 1995, 117, 1076-82.
(17) (a) Holmskov, U.; Fischer, P. B.; Rothmann, A.; Hojrup, P. FEBS
Lett. 1996, 393, 314-6. (b) Adler, P.; Wood, S. J.; Lee, Y. C.; Lee, R. T.;
Petri, W. A. J.; Schnaar, R. L. J. Biol. Chem. 1995, 270, 5164-71.
(
18) Hayashida, O.; Shimizu, C.; Fujimoto, T.; Aoyama, Y. Chem. Lett.
1
998, 1, 13-4.
19) Nicholson, M. W.; Barclay, A. N.; Singer, M. S.; Rosen, S. D.; van
der Merwe, P. A. J. Biol. Chem. 1998, 2, 763-70.
20) (a) Gordon, E. J.; Sanders, W. J.; Kiessling, L. L. Nature 1998,
92, 30-1. (b) Manning, D. D.; Strong, L. E.; Hu, X.; Beck, P. J.; Kiessling,
(
(
3
L. L. Tetrahedron 1997, 53, 11937-52. (b) Manning, D. D.; Hu, X.; Beck,
P.; Kiessling, L. L. J. Am. Chem. Soc. 1997, 119, 3161-2. (c) Schuster,
M. C.; Mortell, K. H.; Hegeman, A. D.; Kiessling, L. L. J. Mol. Catal.
1
997, 116, 209-16.
(
21) For a review discussing the synthesis of bioactive molecules using
ROMP, see: Kiessling, L. L.; Strong, L. E. In BioactiVe Polymers; F u¨ rstner,
A., Ed.; Springer-Verlag: New York, in press.
(22) For other applications of ROMP to the synthesis of carbohydrate-
substituted multivalent ligands, see: (a) Fraser, C.; Grubbs, R. H.
Macromolecules 1995, 28, 7248-55. (b) Nomura, K.; Schrock, R. R.
Macromolecules 1996, 29, 540-5. For reviews describing other approaches
to the synthesis of carbohydrate-substituted materials, see refs 3-5.
(30) Heyse, S.; Vogel, H.; S a¨ nger, M.; Sigrist, H. Protein Sci. 1995, 4,
2532-44. For an overview of the generation and applications of synthetic
surfaces, see: Mrksich, M.; Whitesides, G. M. Annu. ReV. Biophys. Biomol.
Struct. 1996, 25, 55-78.

SPR Measurement of Con A-Ligand Interactions
J. Am. Chem. Soc., Vol. 120, No. 41, 1998 10577
Figure 1. Scheme illustrating control over ligand density on the SPR surface. Liposomes are generated by dissolving POPC and the glycolipid
separately in 50:50 MeOH/CHCl . Combining appropriate volumes of the pure solutions in the desired ratio followed by solvent evaporation and
3
resuspension in aqueous buffers prepares liposomes for injection onto the hydrophobic surface where monolayers are generated.
subtracting high background binding from relatively small,
specific binding events complicated the analysis of the specific
binding values obtained.
Given that Con A binds to dextran, competition studies could
be performed using the underivatized dextran surface. Winzor
and co-workers reported a method for determining the Ka and
concentration of binding sites on the surface for multivalent
interactions using a modified rectangular hyperbolic relation-
3
3
ship.
Using these tools, we analyzed Con A binding to
Figure 2. Scheme for synthesis of the glycolipid (1). (a) Allyl alcohol,
dextran. The number of available surface binding sites and the
Ka of Con A for the dextran surface were evaluated. Subsequent
competition experiments using monovalent saccharides yielded
solution inhibition data. To derive the inhibition constants for
various molecules, the SPR data were analyzed with the solution
competition equation:
+
Dowex-H (39% R). (b) (i) O
3
, 50:50 MeOH/CH
2 2 3
Cl ; (ii) PPh . (c)
NaCNBH
3
, PEtdN, 50:50 MeOH/CHCl
3
(53%, 2 steps).
glucose residues. This surface, which can be used for many
applications, has been designed to minimize nonspecific protein
binding in SPR experiments.31 However, competition studies
using the dextran matrix for immobilization of the carbohydrate
ligand mannose were problematic. When the Con A tetramer
was presented with a dextran surface modified by coupling with
R-C-(ethylamino)-mannose, it showed some increased binding
relative to that obtained with the dextran matrix alone. As
expected, the signal enhancement was small, relative to that
obtained with the unmodified dextran surface. Specifically, with
f ) [I]/([I] + K (1 + F/K ))
i
d
The parameters are defined as: f is fractional inhibition, I is
the inhibitor concentration, Ki is the solution affinity of the
inhibitor for the Con A, F is the concentration of free binding
sites available on the surface, and Kd is the dissociation constant
3
4
of Con A for the surface.
a mannose-substituted surface that gave rise to a signal of 288
response units (RU),32 an injection of Con A (25 µM) elicited
The Ki values determined for monosaccharide inhibition of
the Con A tetramer binding to the dextran surface were
significantly different from the Kd values that had been
a signal of 2799 RU compared to the 2615 RU obtained with
the dextran surface alone. Although Con A has a 3-4-fold
25
higher affinity for mannose over glucose,25 background binding
accurately attained by titration microcalorimetry. Specifically,
the measured Ki values for R-D-methyl-mannopyranoside (RMe-
Man, 2) and R-D-methyl-glucopyranoside (RMeGlc, 3)) (Figure
to the dextran matrix was difficult to dissect from specific Con
A-mannose interactions. Ideally, a control surface, should
demonstrate no affinity for the analyte and simply reflect bulk
refractive index changes in solution. The error associated with
6
2
) inhibition of Con A binding to dextran were 33 ( 10 and
40 ( 30 µM, respectively, data that suggest mannose is 7-fold
more active than glucose. Microcalorimetry experiments indi-
(
(
31) For a review, see: L o¨ fås, S. Pure App. Chem. 1995, 67, 829-34.
32) In surface plasmon resonance analysis, binding of molecules to the
(33) (a) Kalinin, N. L.; Ward, L. D.; Winzor, D. J. Anal.Biochem. 1995,
228, 238-44. (b) Harris, S. J.; Jackson, C. M.; Winzor, D. J. Arch. Biochem.
Biophys. 1995, 316, 20-3.
surface is monitored by refractive index changes. The signal is typically
-
2
expressed in resonance units (1 RU ) 1 pg mm ). For a more detailed
discussion of these measurements, see refs 10-12.
(34) Attie, A. D.; Raines, R. T. J. Chem. Educ. 1995, 72, 119-24.

10578 J. Am. Chem. Soc., Vol. 120, No. 41, 1998
Mann et al.
cated a less dramatic preference between the methyl pyranosides;
Kd values of 130 and 423 µM were obtained for RMeMan and
RMeGlc (Figure 4), respectively, a 3.25-fold preference for
mannose.25 The distorted values observed with the dextran
surface are likely due to the large number of binding sites
available. If the ratio of F, the concentration of free binding
sites, to Kd is greater than 0.1, one of the complexation partners
will be tied up with little available in solution. This situation
35
precludes the accurate determination of even relative affinities.
The F/Kd ratio observed with Con A and the dextran surface
was 0.46, which clearly does not satisfy the experimental
constraints.
Incorporation of Glycolipid 1 To Generate Synthetic
Surfaces. Given the complexities encountered, we reasoned
that a synthetic lipid monolayer could be generated to control
the available ligand density on the surface and, thereby, the
concentration of binding sites.30 In this approach, a gold surface
is modified with a long-chain alkane attached through a thiol
group, upon which a new self-assembled monolayer (SAM) of
defined composition can be generated.1
3,30
We anticipated that
this strategy would identify conditions in which the F to Kd
ratio would be in the desired range, thereby affording a tractable
system for solution competition studies.
Figure 3. Specific binding of Con A tetramer to 10% glycolipid SPR
surface. A. Raw data for injection of Con A over glycolipid and POPC
monolayers. B. Bulk refractive index change subtraction yielding
specific signal data for Con A binding to carbohydrate.
A synthetic glycolipid was designed to bind to Con A. The
target, R-O-ethylaminomannosylphosphatidylethanolamine (1),
displays a terminal mannose residue with an R-anomeric linkage,
the preferred configuration for Con A binding.36 The glycolipid
was altered from 10% to 75%, increased binding to the Con A
tetramer was observed (data not shown). For the binding assays,
the surface with 10% glycolipid was employed since it afforded
the best sensitivity while minimizing the concentration of free
surface binding sites.
The affinity of Con A tetramer for the 10% glycolipid surface
and the concentration of available surface binding sites (F) was
determined by titration with successive injections of 50 µL of
Con A. The concentrations of Con A employed ranged from
was readily constructed using standard transformations and then
incorporated into SAMs. By varying the percentage of synthetic
glycolipid 1 relative to phosphatidylcholine (POPC) in the
injected liposomes, the ligand density was altered, and surface
monolayers composed of POPC and 5, 10, 50, or 75% synthetic
glycolipid were generated. A parallel flow cell, containing a
surface composed solely of unmodified POPC, was used to
assess background bulk refractive index changes in solution and
at the surface (Figure 1). The cell containing POPC only was
used to evaluate the amount and the magnitude of nonspecific
protein binding to the lipid surface. For evaluating Con
A-mannose interactions, responses derived from the POPC cell
were subtracted from those obtained from the mannose-modified
surface to yield values for specific binding of Con A to the
glycolipid surface.
1
7 µM to 0.14 µM, and the solutions were generated by 2-fold
serial dilution. From the raw SPR data, a plot of effective bound
surface concentration versus the effective injected concentration
26
of Con A was generated (Figure 4). Using equilibrium binding
and the rectangular hyperbolic equation to calculate F and Ka
for the binding of the Con A tetramer to the surface, values of
4
-1
0
.97 ( 0.03 µM and (2.7 ( 0.2) × 10 M , respectively, were
obtained. The F/Kd ratio for this system is 0.026, conditions
under which the assumptions in the fractional inhibition equation
are valid.
SPR Experiments
The first objective was to determine the minimum concentra-
tion of glycolipid needed for selective binding of the tetrameric
In the competition assays, inhibition curves were generated
by measuring the binding responses for 500 nM Con A tetramer
in the presence of increasing concentrations of inhibitor.
Fractional inhibition constants were calculated using equilibrium
values generated in the absence of inhibitor (Figure 5). The
inhibition constants (Ki) were generated for RMeMan and
RMeGlc by generating a plot of the fractional inhibition values
obtained versus inhibitor concentration employed. Although
the Ki values that arise from this analysis for RMeMan and
RMeGlc, 92 ( 6 and 290 ( 10 µM respectively, are not
identical to Kd values generated by calorimetry, 130 and 423
µM, the relative affinity values are comparable; a 3.2-fold ratio
from SPR versus 3.25-fold from calorimetry. These results
indicate that this competition assay can reproduce the binding
constants determined by well-tested solution methods.2
Our success in determining monovalent inhibition constants
with this assay system prompted us to investigate the utility of
the method for comparing multivalent ligands. We have been
developing the ring-opening metathesis polymerization (ROMP)
as a method to generate carbohydrate-substituted polymers of
36,37
form of Con A.
As described above, surfaces composed
of 5, 10, 50, and 75% glycolipid-POPC mixtures were
analyzed. At the lowest glycolipid concentration tested (5%),
a weak signal for binding of the Con A tetramer to the surface
was detected. When the glycolipid density of the surface was
augmented to 10%, very good signal-to-noise (1974 RU at 17
µM Con A) and reproducible binding results were obtained.
Under these conditions, the Con A tetramer showed considerably
higher affinity for the surface and displayed a binding profile
with characteristic association, equilibrium, and dissociation
phases. In contrast, the control POPC surface showed only bulk
refractive index changes in signal (Figure 3). Regeneration of
the surface was accomplished by injection of 10 µL 0.1 M H3-
PO4 to remove bound Con A. As the glycolipid surface density
5,36
(
(
35) Horesji, V.; Matousek, V. Mol. Immunol. 1985, 22, 125-33.
36) Bittiger, H.; Schnebli, H. P. ConcanaValin A as a Tool; John Wiley
and Sons: London, 1976.
(37) Derewenda, Z.; Yariv, J.; Helliwell, J. R.; Kalb, A. J.; Dodson, E.
J.; Papiz, M. Z.; Wan, T.; Campbell, J. EMBO J. 1989, 8, 2189-93.

SPR Measurement of Con A-Ligand Interactions
J. Am. Chem. Soc., Vol. 120, No. 41, 1998 10579
Figure 6. Inhibitors used in SPR experiments. 2 and 3: monovalent
ligands for Con A. 4 and 5: neoglycopolymers of different lengths
used to explore structure/function relationships in multivalent binding.
Figure 4. Titration of Con A tetramer on glycolipid surface. A. Data
used to determine the K
concentration of available binding sites. B. Evaluation of the results
by the rectangular hyperbolic equation to determine K and F.
a
of Con A binding to the surface, and F, the
a
Figure 7. Polymer inhibition of Con A tetramer binding to glycolipid.
K values are normalized to RMeMan (2). Polymer concentrations were
i
calculated on a per residue basis. Data from no preincubation or 1 h
incubation times of Con A with polymers before injection onto the
glycolipid surface are reported.
6
) had been generated in conjunction with these studies.38 To
investigate the abilities of these agents to inhibit the binding of
Con A to the glycolipid surface, solution competition experi-
ments with the neoglycopolymers were carried out in a manner
analogous to those performed with RMeMan 2 and RMeGlc 3.
Polymer length was estimated by NMR integration analysis
referencing internal double bond protons to the terminal aromatic
39
protons. For these substrates, the reported inhibitor concentra-
tions were calculated on a saccharide residue basis. The
resulting Ki values, therefore, do not report on the molar
inhibitory constants of the polymers, but rather the average of
the saccharide units within the polymer (Figure 7).
The inhibition data from the polymers reveal that the relative
potencies of the mannose-substituted polymers increase with
their length. For example, the 10mer 4a is 3-fold less active
than the 25mer 4b, a trend that is not continued when the
polymer length is extended to 143 residues, as in 4c. As
expected, no inhibition was observed with the galactose 25mer
5
because Con A does not bind galactose. This result provides
further evidence that the Con A interaction with the surface is
specifically occurring via the mannose-substituted glycolipid;
only saccharide ligands that bind Con A inhibit surface binding.
Figure 5. Inhibition of 500 nM Con A tetramer binding to the
glycolipid surface. A. Inhibition curves for 0, 0.1, 0.2, 1, 2, 5, and 10
mM RMeGlc (3). B. Fractional inhibition curves for RMeMan (2) and
RMeGlc (3) inhibition of Con A binding.
(
38) Kanai, M.; Mortell, K. H.; Kiessling, L. L. J. Am. Chem. Soc. 1997,
19, 9931-2.
39) Molecular masses calculated in this manner correspond well to those
1
(
defined length,2
0-23
which have been termed neoglycopolymers.
determined by gel permeation chromatography: Kanai, M., unpublished
results.
A series of mannose- and galactose-substituted materials (Figure

10580 J. Am. Chem. Soc., Vol. 120, No. 41, 1998
Mann et al.
These data also rule out nonspecific inhibition by the polymer
backbone as a mechanism of action of the mannose derivatives.
Because the polymers are much larger than the low molecular
weight monosaccharides investigated previously, the kinetics
of binding might be a factor in their observed inhibitory
potencies. To examine this issue, the polymers were incubated
with Con A for 1 h, prior to their use in SPR competition studies.
In these experiments, an increase in the relative potency, from
5-fold up to 40-fold, was observed when compared to their
activities with no preincubation. Significantly, the same
dependence on polymer length was obtained, with no dramatic
increase in potency occurring for materials with lengths beyond
the 25mer.
Discussion
The self-assembled glycolipid monolayer provided an excel-
lent surface for assessing the affinity of Con A for monovalent
ligands in solution via competition experiments. The sensitivity
achieved by this method enhances the detectable affinity range
observable by SPR. Although Con A binds only weakly to
monosaccharides, the relative solution affinities of Con A for
these compounds could be measured by this method. This
assay, therefore, provides a method for rapidly screening low
molecular weight compounds for inhibition of protein-
carbohydrate interactions. In addition, the relative activities of
the multivalent inhibitors for the Con A tetramer could be
assessed conveniently and rapidly.
Figure 8. Multivalent modes of binding. A. Binding enhancement due
to simultaneous spanning of two binding sites. B. Enhancement of
binding due to increased local concentration of available ligand. C.
Both modes of inhibition illustrated in the model system reported with
Con A tetramer binding to neoglycopolymers.
minor increases in activity (Figure 7). Thus, these data suggest
that the neoglycopolymers act most effectively when their length
permits concurrent multiple binding site interactions. These
overall findings with regard to multivalent ligand activity using
the SPR competition assay are consistent with our studies of
The investigations involving the neoglycopolymers provide
further insight into multivalent ligand binding to proteins,
especially as it applies to the interplay between multivalent
ligand length and inhibitory potency. The Con A tetramer
presents two saccharide binding sites on each face, and the
orientation of these two sites allows Con A to engage in
38
cell agglutination.
Although the observed trends in inhibitory potency for various
molecules are similar, the activities of compounds in the cell
38
agglutination assay span a dynamic range broader than those
obtained from SPR. For example, in the SPR assay, the
maximum increase in inhibitory potency for the most potent
multivalent mannose derivative is approximately 40-fold over
that of monovalent RMeMan. Augmentations of approximately
40,41
multivalent interactions.
The distance between the two
relevant saccharide binding sites within the Con A tetramer is
approximately 65 Å, as determined from X-ray structural
analysis.3
6,37
The neoglycopolymer length needed to place
1
000-fold were found in the cell agglutination study. It is not
saccharide residues in each of the binding sites can be
approximated using molecular mechanics calculations. Polymer
models of stereochemically homogeneous linkages were gener-
ated to provide insight into the geometries that can be adopted.
The synthetic materials are composed of a mixture of cis- and
trans-alkene isomers and isotactic and syndiotactic stereo-
isomers; however, by focusing on the limiting cases, estimates
of polymer lengths could be generated. The least extended
structure is the all-cis-syndiotactic isomer, in which the rise per
residue is approximately 1.9 Å. The most extended structure
is the all-trans-syndiotactic isomer, with an average separation
of approximately 4.5 Å between residues. On the basis of these
calculations, the minimum polymer length that could span the
surprising that large differences in relative potencies may occur
between two distinctly different assays, especially when those
assays involve multivalent binding interactions. The observed
magnitude of enhancement when comparing a monovalent to a
multivalent ligand will depend on the conditions under which
the ligand must function. For example, the activities of
compounds measured in static assays may be different than those
20c
46,47
measured under flow.
Under the SPR assay conditions, the mechanism of action of
the polymers may not involve simultaneous placement of
saccharide epitopes in the Con A binding sites. Specifically,
the activities of the polymers could be attributed to their abilities
(42) For studies exploring cross-linking of concanavalin A in the solid
6
3
5 Å gap between the two binding sites would be between a
5mer cis-syndiotactic polymer or a 15mer trans-syndiotactic
state, see: (a) Mandal, D. K.; Brewer, C. F. Biochemistry 1992, 31, 12602-
9
. (b) Mandal, D. K.; Brewer, C. F. Biochemistry 1993, 32, 5117-20.
polymer (Figure 8). These data suggest that neoglycopolymers
of 25 or more residues can interact with Con A in a divalent
manner.
(43) Several studies have identified multivalent ligands that are more
potent inhibitors of lectin binding to cells than they are of lectin-saccharide
binding in solution. The authors have proposed that the multivalent ligands
can cross-link the lectins in the cellular environment. For some representative
examples, see: (a) Lee, Y. C. In Binding modes of mammalian hepatic
Gal/GalNAc receptors; Lee, Y. C., Ed.; John Wiley & Sons: Chicester,
U.K., 1989; Vol. 145, pp 80-95. (b) Glick, G. D.; Toogood, P. L.; Wiley:
D. C.; Skehel, J. J.; Knowles, J. R. J. Biol. Chem. 1991, 266, 23660-9.
The results from SPR detection indicate that the most potent
neoglycopolymers are those that can simultaneously bridge two
40,41
saccharide binding sites within the Con A tetramer.
More-
over, once multiple binding sites can be spanned by a single
binding partner, polymers with increased length exhibit only
(44) Quesenberry, M. S.; Lee, R. T.; Lee, Y. C. Biochemistry 1997, 36,
2724-32.
(45) Lee, R. T.; Ichikawa, Y.; Kawasaki, T.; Drickamer, K.; Lee, Y. C.
(
40) For example, succinylated Con A which exists as a dimer with
Arch. Biochem. Biophys. 1992, 299, 129-36.
(46) Alon, R.; Feizi, T.; Yuen, C.-T.; Fuhlbrigge, R. C.; Springer, T. A.
J. Immunol. 1995, 154, 5356-5366.
(47) Sanders, W. J.; Gordon, E. J.; Alon, R.; Kiessling, L. L., unpublished
results.
binding sites oriented in opposite directions does not engage in multivalent
interactions.
(41) For a discussion of the chelate effect, see: Page, M. I.; Jencks, W.
P. Proc. Natl. Acad. Sci. U.S.A. 1971, 68, 1678-83.

SPR Measurement of Con A-Ligand Interactions
J. Am. Chem. Soc., Vol. 120, No. 41, 1998 10581
to noncovalently cross-link Con A tetramers, as has been
observed for other multivalent ligands.42 With this mechanism
of action, however, it would be expected that the inhibitory
potencies of the neoglycopolymers would continue to increase
with molecular mass with longer polymers exhibiting ever
increasing potencies. Such an effect is not observed (Figure
in the magnitudes of the potencies observed. An important
conclusion from our results is that both assays provide the same
relative ranking of the multivalent ligands and reproduce the
38
trends in their binding potencies. Thus, the SPR assay is useful
for measuring the relative affinities of molecules within a class
of compounds (monovalent or multivalent). Our studies reveal
that SPR is an effective method for screening the activities of
a series of low molecular weight, monovalent ligands. Signifi-
cantly, competition assays such as the one reported here can be
used to rank inhibitors of weak protein-saccharide interactions.
This information will illuminate the molecular features important
for complexation and will provide a basis for optimizing
inhibitor structure. Moreover, the SPR competition assay can
also be used to order inhibitor potencies within series of
multidentate ligands, providing a method to rapidly identify
potent ligands from this class of compounds.
7
). Moreover, the most active multivalent compounds in both
assays are those with the ability to span the requisite distance
to occupy two Con A saccharide binding sites.
The differences in the results of the two assay systems
illuminate important features of multivalent binding part-
ners.2
7,41-45
For example, the kinetics of the relevant interac-
tions may lead to large differences between the two assays. In
the cell agglutination assay, the assay is conducted over a period
of several hours. With the optical biosensor, equilibrium must
be established in seconds to minutes for the competition results
to be accurate. With the SPR assay, an inherent difficulty in
performing competition experiments under flow conditions is
that the rate of dissociation of the analyte, Con A, from the
glycolipid surface depends on the surface composition and local
structure. Because Con A is multivalent, it has a high
probability of “rebinding” to the surface, which renders the rate
of dissociation of the lectin slow on the time scale of the
competition experiments (Figure 3). Thus, it is difficult to
establish a binding equilibrium among the three components
of the mixture, Con A, the glycolipid surface, and the inhibitor
tested under the conditions of flow. Rapid equilibrium can be
established with monosaccharides but is clearly not reached
when the neoglycopolymers are tested in our assay system
Summary
Not all physiologically and medically relevant receptor-
ligand interactions occur with high affinities. Saccharides
constitute one class of ligands that often functions by associating
weakly with their target proteins. Methods to rapidly and
accurately monitor low affinity interactions can facilitate the
development of structure and function relationships in such
systems and lay the groundwork for the identification of
inhibitors. To address this issue, an artificial ligand-containing
bilayer was generated, and a surface plasmon resonance assay
was developed to assay the ability of various inhibitors to block
protein binding to the synthetic bilayer. The system investigated
involved the binding of the lectin Con A to a bilayer containing
synthetic mannose-substituted glycolipids. Results with known
monovalent Con A ligands reveal the effectiveness of SPR for
these purposes. Significantly, this method can be used to rapidly
evaluate the inhibitory potencies of a series of low molecular
weight ligands.
(Figure 7). Other studies of protein-carbohydrate interactions
suggest that the evaluation of binding events under conditions
of flow affords results distinct from those determined under
static conditions.4
6,47
Interestingly, the Whitesides group has
shown that two different static assays afford similar results in
the evaluation of the inhibitory potencies of polyvalent sialic
48
acid-substituted ligands toward influenza virus hemagglutinin.
In addition to its effectiveness for investigating low affinity
interactions, the SPR assay can provide insight into the relative
ranking of a series of multivalent ligands. To investigate its
utility for this purpose, the inhibitory potencies of a series of
neoglycopolymers composed of various lengths were assessed.
As was observed previously in cell agglutination studies, the
most potent multivalent derivatives were those that could
simultaneously occupy two saccharide binding sites within the
Con A tetramer. The increases in potencies for the multivalent
ligands that were observed in the cell agglutination assay were
dramatic relative to those seen with SPR. These results
underscore the differences that can arise when evaluating
interactions with an artificial surface relative to those observed
with cells.
The assays employed occur under static binding conditions on
time scales which allow for equilibrium between inhibitor and
detectable binding partner to be reached. The dynamic nature
of the SPR assay requires that equilibrium be rapidly attained,
which could lead to the potency differences observed here.
The organization and presentation of ligands on the very
different surfaces, the artificial glycolipid bilayer, and the red
blood cell used in the two assays may also be manifested in
the divergent results. For example, the density of ligands
displayed by each surface is likely to be very different. In the
cell agglutination experiment, Con A binds to surface glyco-
proteins with mannose residues, while in the SPR assay, ConA
binds to artificial glycolipid-containing bilayers. The rates of
Con A dissociation from these surfaces should differ, and these
rates can influence the outcome of the competition experiments,
as described above. Another important feature that separates
hemagglutination and artificial surface binding inhibition is the
mechanism by which inhibition can be achieved. Since Con A
is a tetramer, agglutination will be effectively inhibited by
occupying two Con A subsites, but such a binding event will
not necessarily inhibit surface binding to the other face of the
tetramer (Figure 8). In addition, if polymer-bound Con A binds
to the glycolipid surface, dissociation will also be slightly slowed
by the increase in molecular mass.
Experimental Materials and Methods
General Methods. All materials were obtained from commercial
suppliers and used as provided. Solvents and allyl alcohol were distilled
by standard protocols. All reactions were monitored by thin-layer
chromatography (TLC) on 0.25-mm precoated Merck silica gel 60 F254
and visualized with phosphomolybdic acid or p-anisaldehyde stains.
Flash column chromatography was performed on Merck silica gel 60
(230-400 mesh). 1H and C spectra were recorded on a Bruker AC-
13
3
00 or Varian Unity 500 spectrometer and are referenced solvent peaks
1
13
3
(CDCl , H δ 7.24, C δ 77.0). Mass spectrum data was collected on
Thus, the inherent differences in the time scales of the assays,
the structures of the surfaces, and the nature of agglutination
versus surface binding is likely to account for the discrepancies
a Bruker MALDI-TOF instrument using R-cyano-4-hydroxysuccinamic
acid matrix. SPR measurements were performed on a BIAcore 2000
operated using the Version 1.3 software. Buffers were sterile-filtered
and deoxygenated. Con A was purchased from Vector Laboratories,
Inc., Burlingame, CA.
(48) Sigal, G. B.; Mammen, M.; Dahmann, G.; Whitesides, G. M. J.
Am. Chem. Soc. 1996, 118, 8(16), 3789-3800.

1
0582 J. Am. Chem. Soc., Vol. 120, No. 41, 1998
Mann et al.
containing 0.2%(w/v) calcium chloride) yielded 1 (15.0 mg, 53%): R
Preparation of Mannosylated Dextran Matrix. The CM5 chip
f
1
carboxylated dextran was activated by injection of a mixture of N-ethyl-
N′-(diethylaminopropyl)-carbodiimide (EDC) and N-hydroxysuccin-
imide (NHS) (120 µL, 200 mM EDC, 50 mM NHS, 20 µL/minute)
dispensed by the BIAcore instrument. R-C-Ethylaminomannose (0.100
mg/mL in 10 mM HEPES buffer, pH 6.0) was injected into flowcell 1
3 3
) 0.10 (2:1 chloroform/methanol); H (500 MHz, 50:50 CD OD/CDCl )
δ 5.26-5.20 (1H, m), δ 4.42 (2H, dd, J ) 12.0, 3.0 Hz), δ 4.17 (1H,
dd, J ) 12.0, 6.5), δ 4.80-3.46 (14H, m), δ 3.30-2.70 (4H, m), δ
2.31 (4H, q, J ) 7.5), δ 1.70-1.55 (4H, m), δ 1.40-1.20 (52H, m), δ
0.869 (6H, t, J ) 7.0); 13C (500 MHz, 50:50 CD
3 3
OD/CDCl ) δ 180.4,
(120 s contact time) and flowcell 2 (30 s) and allowed an additional
174.5, 147.1, 101.0, 100.8, 74.0, 73.8, 73.6, 71.7, 71.5, 71.3, 71.1, 68.1,
68.1, 67.9, 66.2, 65.3, 65.1, 64.2, 63.6, 63.2, 62.2, 62.1, 57.6, 57.2,
55.9, 34.7, 34.6, 32.5, 30.2, 30.1, 29.9, 29.7, 25.5, 23.2, 14.2. MALDI
3
0 s to react. Unreacted, activated carboxyl groups were capped by
the injection of ethanolamine (120 µL, 1 M pH 8.5) yielding an increase
of 278 and 288 RU in flowcells 1 and 2, respectively.
m/z M calcd for C45H87NO14P 896.6 g/mol, observed 896.5.
SPR on Dextran Matrix. Con A (2.5 mg, 96 nmol) was dissolved
in sample buffer (1.0 mL, 20 mM HEPES, pH 7.0, 90 mM NaCl)
overnight at 4 °C and syringe-filtered (0.22 µm). Final concentration
was determined by measuring the absorbance at 280 nm (A280 ) 1.37
SPR on Glycolipid. An HPA chip was perfused overnight with
degassed water (5 µL/min), the hydrophobic surface was washed with
octyl glucoside (25 µL, 40 mM). Liposome mixtures of defined
glycolipid/phosphatidylcholine (POPC) concentration were generated
by mixing pure samples dissolved in 50:50 chloroform/methanol in
the desired molar ratios. The solvent was evaporated under an argon
stream, and liposomes were generated by resuspension in sample buffer
(0.5 M in 20 mM HEPES, pH 7.0, 90 mM NaCl). The liposome
suspension was injected (300 µL) onto the lipid surface leading to an
increase of approximately 1500 Response Units over initial background.
After 30 min of washing at 5 µL/min, the surface was purged at 1000
µL/min for 5 min to remove weakly attached lipid vesicles.
×
[mg/mL Con A]). Surface density titration was carried out by
successive injections of Con A (50 µL), allowing 500 s for dissociation,
followed by regeneration (10 µL, 0.1 M H PO4). The titration range
covered 12.5 to 0.39 µM by 2-fold dilutions. F and K were determined
3
a
3
3
as reported by Kalinin.
Competition experiments with RMeMan (5) and RMeGlc (6) were
carried out by running a method supplied by the BIAcore software,
version 1.3. Con A tetramer (100 nM) was mixed with inhibitor and
this mixture (50 µL) was injected (10 µL/minute). Inhibitor concentra-
tions of 10.0, 8.00, 6.00, 4.00, 2.00, 1.00, 0.500, and 0.100 mM were
tested. Bound Con A response values were assessed during the
After changing the solvent to degassed sample buffer, a titration
was carried out. Injected concentrations ranged from 17 µM to 0.14
µM Con A tetramer by differing 2-fold dilutions (50 µL KINJECT, 10
equilibrium binding portion of the curve (295 s after injection). K
values were determined by fitting the data to the fractional inhibition
equation: f ) [I]/([I] + K (1 + F/K )).
i
a
µL/minute). F and K for Con A tetramer on the glycolipid surface
were evaluated as reported for the dextran matrix. Solution inhibition
experiments were carried out by mixing equal volumes of Con A
tetramer (1 µM) and double the concentration of inhibitor to be tested.
i
d
Synthesis of Glycolipid 1. D-Mannose (1.00 g, 5.10 mmol) was
dissolved in allyl alcohol (10.0 mL, 160 mmol). Freshly activated acidic
Dowex resin (0.504 g, 50 × 8-200 mesh) was added and the mixture
refluxed under nitrogen for 1.5 h. Column chromatography purification
5
0 µL of the resultant sample mixture (0.5 µM Con A tetramer/1 ×
inhibitor) was injected (10 µL/minute) and equilibrium binding response
values were taken at equilibrium binding (260 s postinjection). The
inhibitory potencies of the polymers were generated using either a direct
injection or a 1 h preincubation after mixing the Con A and inhibitor
prior to injection of the competition mixture. Bulk refractive index
data was generated on a parallel 100% POPC coated flow cell.
(silica, 30% 9:4:2 ethyl acetate/n-propyl alcohol/water in ether) yielded
the R-anomer (0.446 g, 39%). The allylated pyranoside (7.0 mg, 0.032
mmol) was dissolved in methanol (0.50 mL) and dichloromethane (0.50
mL) and cooled to -78 °C under a nitrogen atmosphere. Ozone was
bubbled through the mixture for 5 min. Polymer-bound triphenyl
phosphine (20.0 mg, 0.0636 mmol) was added, and the reaction was
allowed to warm to room temperature and stir for 3 h. The triph-
enylphosphine oxide was removed by filtration over a small plug of
Celite, followed by washing with methanol (3.0 mL), to yield the desired
aldehyde (7.0 mg, 0.032 mmol, 100%). The solvent was evaporated,
and the resultant solid was dissolved in dimethyl sulfoxide (100 µL).
Methanol (500 µL) was added, and this mixture was added to
phosphatidylethanolamine (22.0 mg, 0.0319 mmol) dissolved in
chloroform (0.50 mL) and methanol (0.25 mL). The reducing mixture
Acknowledgment. This research was supported by the NIH
(GM55984) and the Mizutani Foundation. L.L.K. thanks the
American Cancer Society, the NSF (NYI program), the Dreyfus
Foundation, Zeneca Pharmaceuticals, and the Sloan Foundation
for support. Biacore data were obtained at the University of
Wisconsin-Madison Biophysics Instrumentation Facility (BIF),
which is supported by the University of Wisconsin-Madison
and grant BIR-9512577 (NSF). We thank Dr. Chris Whalen
(1% (w/v) sodium cyanoborohydride, 0.1% (v/v) acetic acid) in
(BIAcore, AB) and Dr. Darrell McCaslin (UW-Madison, BIF)
chloroform (0.25 mL) and methanol (0.25 mL) was then added. The
resulting mixture was allowed to stir at room temperature for 30 h.
Purification by flash chromatography (silica, 65:35 chloroform/methanol
for helpful conversations.
JA9818506