High-affinity pentavalent ligands of Escherichia coli heat-labile enterotoxin by modular structure-based design [10]

Full text of DOI:10.1021/ja993388a

J. Am. Chem. Soc. 2000, 122, 2663-2664
2663
Scheme 1
High-Affinity Pentavalent Ligands of Escherichia coli
Heat-Labile Enterotoxin by Modular Structure-Based
Design
Erkang Fan,* Zhongsheng Zhang, Wendy E. Minke,
Zheng Hou, Christophe L. M. J. Verlinde, and Wim G. J. Hol†
Department of Biological Structure
Biomolecular Structure Center and
Howard Hughes Medical Institute
UniVersity of Washington, Box 357742
Seattle, Washington 98195
“core” that can adopt a 5-fold symmetric configuration and
provides a foundation for the generation of structural comple-
mentarity for the overall pentavalent ligand to LT, “linkers” that
project in the direction of the receptor binding sites, and “fingers”
that fit snugly into the binding sites (Scheme 1). With efficient
chemistry to connect these modules, assembly of the large
pentavalent ligands is synthetically feasible, as will be shown.
In this report, we demonstrate the power of a modular synthesis
procedure, which allowed us to explore in detail the effects of
linker length on affinity. For the core, we chose acylated
pentacyclen 5. Force-field calculations show that this molecule
can adopt a conformation close to 5-fold symmetry. We used 1-â-
amidated D-galactose 1 as the finger. D-galactose is a terminal
sugar unit of LT’s natural receptor GM1. It interacts very
specifically with the toxin via defined hydrogen bonds and a
carbohydrate against tryptophan stacking.7a Galactose and its
derivatives with substitutions at C1 have been observed to bind
in the same manner in LT’s receptor binding site.7 Thus, 1 would
be expected to be a good mimic of galactose while providing
functional groups for full ligand assembly.
The success of this design, however, depends critically on
finding suitable linker modules. Rigid linkers complementary to
the protein surface8 would be ideal but are difficult to achieve in
one design step. Here we present a solution to this problem, based
on readily available flexible linkers. This raises the following
questions: can a large gain in affinity be obtained with flexible
linkers, and how long do the linkers need to be? To answer these
questions, we have chosen to span a large range of linker lengths
using the commercially available 4,7,10-trioxa-1,13-tridecanedi-
amine 3 as the basic unit of the linkers. Longer linkers in our
pentavalent ligands consist of up to four units of 3.
The modular synthesis of our full ligands is shown in Scheme
2. Although each long linker could be synthesized separately from
3, it is more economical to perform the stepwise coupling of each
unit of 3 to the core-linker assembly. As an alternative to the
HPLC purification of each reaction intermediate, we also devel-
oped purification protocols based on C18 Sep-Pak cartridges9
which can handle far greater sample load than a typical research-
lab preparative HPLC setup. The squaric acid diester mediated
coupling reaction10 was very clean and efficient. There was no
partially derivatized product detectable at each reaction step.
Typical isolated yields for each coupling product were around
80∼95%. To the best of our knowledge, this represents one of
the first efficient syntheses of pure single-species protein ligands
with large molecular weights (4-8 kDa).
ReceiVed September 20, 1999
High-affinity protein ligands have wide applications in disease
diagnosis, prevention, and treatment. A promising strategy to
arrive at such ligands is the creation of multivalent compounds
that bind to a multivalent target.1 Current approaches include the
synthesis of bivalent ligands, in which two copies of a ligand are
joined by a flexible linker;2 and more generally, the attachment
of a larger number of monovalent ligands onto a selected
backbone. Examples of backbones include polymers/oligomers,3
membranes,4 and dendrimers.5 In these approaches structure-based
information on the spatial arrangement of the target’s multiple
binding sites does not usually enter the design process. Therefore,
the obtained ligands may not be ideal for maximizing the
interaction with the target. Here, we present a novel approach
toward high-affinity multivalent ligands: a modular design that
incorporates structural information of the multiple target sites.
Our work focuses on an ideal target model: the heat-labile
enterotoxin (LT) from Escherichia coli. LT, a close relative of
cholera toxin, is a member of the AB5 family of bacterial toxins
which also includes shiga toxin, shiga-like toxin, and pertussis
toxin.6 The symmetrical arrangement of the five B subunits of
LT presents five identical carbohydrate binding sites that recognize
ganglioside GM1 headgroups protruding from cells of the
gastrointestinal lumen. This LT-carbohydrate interaction has been
determined in atomic detail,7 and provides the basis for the
structure-based design of pentavalent ligands.
The large dimensions of LT make the design and synthesis of
a structurally complementary pentavalent ligand a major chal-
lenge. As shown in Scheme 1, distances between the toxin’s
nonadjacent binding sites are 45 Å. In our modular design, the
large pentavalent ligand is divided into three modules: a semirigid
† Howard Hughes Medical Institute, University of Washington.
(1) Mammen, M.; Choi, S.-K.; Whitesides, G. M. Angew. Chem., Int. Ed.
1998, 37, 2754-2794.
(2) (a) Kramer, R. H.; Karpen, J. W. Nature 1998, 395, 710-713. (b) Profit,
A. A.; Lee, T. R.; Lawrence, D. S. J. Am. Chem. Soc. 1999, 121, 280-283.
(c) Glick, G. D.; Toogood, P. L.; Wiley: D. C.; Skehel, J. J.; Knowles, J. R.
J. Biol. Chem. 1991, 266, 23660-23669.
(3) (a) Kanai, M.; Mortell, K. H.; Kiessling, L. L. J. Am. Chem. Soc. 1997,
119, 9931-9932. (b) Mammen, M.; Dahmann, G.; Whitesides, G. M. J. Med.
Chem. 1995, 38, 4179-4190. (c) Connolly, D. T.; Townsend, R. R.;
Kawaguchi, K.; Bell, W. R.; Lee, Y. C. J. Biol. Chem. 1982, 257, 939-945.
(d) Lee, T. C. FASEB J. 1992, 6, 3193-3200.
(4) (a) Kingery-Wood, J. E.; Williams, K. W.; Sigal, G. B.; Whitesides,
G. M. J. Am. Chem. Soc. 1992, 114, 7303-7305. (b) Spevak, W.; Nagy, J.
O.; Charych, D. H.; Schaefer, M. E.; Gilbert, J. H. J. Am. Chem. Soc. 1993,
115, 1146-1147.
(5) (a) Thompson, J. P.; Schengrund, C. L. Glycoconjugate J. 1997, 14,
837-845. (b) Roy, R. Curr. Opin. Struct. Biol. 1996, 6, 692-702. (c) Kitov,
P. I.; Sadowska, J. M.; Bundle, D. R.; Read, R. R.; Armstrong G. D. Abstracts
of the XV Interactional Symposium on Glycoconjugates, Tokyo, 1999; pp S44.
(6) Merritt, E. A.; Hol, W. G. J. Curr. Opin. Struct. Biol. 1995, 5, 165-
171.
(7) (a) Sixma, T. K.; Pronk, S. E.; Kalk, K. H.; Van Zanten, B. A.; Berghuis,
A. M.; Hol, W. G. J. Nature 1992, 355, 561-564. (b) Van den Akker, F.;
Feil, I. K.; Roach, C.; Platas, A. A.; Merritt, E. A.; Hol, W. G. J. Protein Sci.
1997, 6, 2644-2649. (c) Merritt, E. A.; Sarfaty, S.; Feil, I. K.; Hol, W. G. J.
Structure 1997, 5, 1485-1499.
After obtaining a series of pentavalent ligands with various
linker lengths (10-13), we tested their ability to inhibit the
binding of LT B pentamer (LT-B5) to ganglioside, using the
identical ELISA protocol published previously11 (Table 1). Each
(8) Park, H. S.; Lin, Q.; Hamilton, A. D. J. Am. Chem. Soc. 1999, 121,
8-13.
(9) Waters Sep-Pak cartridge application bibliography, 5th ed.; McDonald,
P. D., Ed.; Waters: Milford, MA, 1991.
(10) Tietze, L. F.; Arlt, M.; Beller, M.; Glusenkamp, K.-H.; Jahde, E.;
Rajewsky, F. Chem. Ber. 1991, 124, 1215-1221.
10.1021/ja993388a CCC: $19.00 © 2000 American Chemical Society
Published on Web 03/03/2000

2664 J. Am. Chem. Soc., Vol. 122, No. 11, 2000
Communications to the Editor
Scheme 2
Figure 1. Plot of IC50’s versus the dimensions of ligands.
note that in this first systematic study of the effects of flexible
linker length on pentavalent ligand affinity, our results agree very
well with the notion brought up by Kramer and Karpen2a for
flexible bivalent ligands. That is: in order to obtain a large gain
in affinity, the ligand dimensions should match that of the binding
site distribution in the target (Scheme 1). Most importantly, this
match should not be based on the extended conformation of the
flexible linkers, rather, the linkers’ effectiVe lengths should be
considered. As a guideline for determining the effective dimen-
sions of our pentavalent ligands, we followed the equation used
by Kramer and Karpen, which is based on a polymer model
described by Knoll and Hermans.13 Hence, the effective dimen-
sions of our ligands are taken to be proportional to the square
root of the molecular weights between two fingers. As shown in
Figure 1, even though all ligands’ extended dimensions are far
greater than the distance between nonadjacent binding sites in
LT, the effective dimensions of ligands 10-12 are less than the
nonadjacent binding site distance in LT. Therefore, we see a
steady drop in IC50’s as the ligand’s effective dimensions increase
in ligands 10-13. In ligand 13, where the ligand’s effect
dimensions have the best match of LT’s binding site distribution
in this series of ligands, we obtained the largest gain in affinity
to LT over galactose. Although it would be interesting to see the
affinities of ligands with effective linker lengths substantially
larger than the distances between the target binding sites, the low
solubilities of the synthetic intermediates prevented us from
obtaining such ligands. It is, however, worth noting that according
to the findings of Kramer and Karpen,2a ligands with too long a
linker length would have less affinity toward the target protein
than ligands whose effective dimensions match the distribution
of binding sites on the target protein.
Table 1
ligand
IC50 (µM)
gain over galactose
galactose
14
10
11
12
13
GM1-OS
58000 ( 8000
5000 ( 200
242 ( 91
16 ( 8
6 ( 4
0.56 ( 0.06
0.01 ( 0.01
1
11
240
3600
10 000
104 000
5 800 000
IC50 was the standard-deviation-weighted average of two or more
independent quadruplicate determinations.
The results in Table 1 clearly show that our structure-based
design of pentavalent ligands leads to very significant affinity
gains compared to the monovalent ligand. The best pentavalent
ligand 13 shows an IC50 that is 105-fold better than galactose,
the molecular moiety mostly responsible for the affinity of our
fingers to LT. The IC50 of 13 approaches that of GM1-OS, the
oligosaccharide portion of LT’s natural receptor ganglioside
GM1.11 To further dissect the effect of pentavalent linkage rather
than that of linker structure, we synthesized compound 14
consisting of both the finger and the linker modules. Compared
to 14, ligand 13 shows more than 104-fold gain in affinity, or
2000-fold on a valency-corrected basis. In comparison, recently
reported GM1-OS derivatized dendrimers only exhibited a few
100-fold affinity increase per finger compared to GM1-OS by
itself,5a which underlines the efficiency of our approach. To
demonstrate that the improvements in IC50’s by our pentavalent
ligands over galactose is most likely due to the formation of a
In summary, our modular approach has allowed for efficient
synthesis of large molecular weight protein ligands and, for the
first time, a systematic study of the effects of flexible-linker
lengths on the affinities of multivalent ligands. The design of
multivalent ligands on the basis of structural information appears
more powerful in gaining ligand affinity than conventional
approaches based on structurally less well-defined supports.
Further study of multivalent ligands based on our approach, such
as using higher affinity fingers11 and incorporating protein surface
recognition8 into linkers, would provide even more insight into
multivalent protein-ligand interactions.
Acknowledgment. We are grateful to Dr. Tim Hirst for providing
antibodies used in the ELISA and Dr. Ethan Merritt for helpful
discussions. This work is supported by the School of Medicine (to E.F.),
and grants from the National Institute of Health (GM54618 to C.V. and
AI34501 to W.H.).
1:1 complex in solution, rather than due to the effect of cross-
linking target protein in solution,2c,
we have performed
3c, 5c
Supporting Information Available: Preparation of compounds and
DLS experiments (PDF). This material is available free of charge via
the Internet at http://pubs.acs.org.
dynamic light scattering (DLS) studies on ligand-protein com-
plexes. Solutions of LT-B5 at micromolar concentrations in the
presence of 0.5, 1, or 2 equiv of ligand 13 showed no evidence
of aggregation.12 This clearly ruled out the cross-linking effect
by our pentavalent ligand and is supportive for the formation of
a 1:1 complex in solution between 13 and LT-B5.
JA993388A
(12) The DLS-measured molecular weights for LT-B5 in solution alone at
3.9 and 7.8 µM are 54 and 52 kD, respectively (theoretical value: 59 kD). In
the presence of various amounts of 13, the measured molecular weights varied
from 53 to 57 kD. As a positive control for protein aggregation, a solution of
LT-B5 at 7.8 µM in the presence of the aggregating reagent chlorophenolred-
â-D-galactopyranoside showed a molecular weight of >400 kD. The detailed
experimental data is provided in the Supporting Information.
In addition to demonstrating a large affinity gain by our
pentavalent ligands over the monovalent ligand, it is exciting to
(11) Minke, W. E.; Roach, C.; Hol, W. G. J.; Verlinde, C. L. M. J.
Biochemistry 1999, 38, 5684-5692.
(13) Knoll, D.; Hermans, J. J. Biol. Chem. 1983, 258, 5710-5715.