Structural basis of multivalent binding to wheat germ agglutinin

Published on Web 06/07/2010

Structural Basis of Multivalent Binding to Wheat Germ

Agglutinin

David Schwefel,^†Caroline Maierhofer,^‡Johannes G. Beck,^‡Sonja Seeberger,^‡,§

Kay Diederichs,^†Heiko M. Mo¨ller,*^,‡Wolfram Welte,*^,†and Valentin Wittmann*^,‡

Departments of Chemistry and Biology, UniVersita¨t Konstanz, UniVersita¨tsstr. 10,

78457 Konstanz, Germany

Received February 25, 2010; E-mail: Mail@valentin-wittmann.de; Wolfram.Welte@uni-konstanz.de;

Heiko.Moeller@uni-konstanz.de

Abstract: The inhibition of carbohydrate-protein interactions by tailored multivalent ligands is a powerful

strategy for the treatment of many human diseases. Crucial for the success of this approach is an

understanding of the molecular mechanisms as to how a binding enhancement of a multivalent ligand is

achieved. We have synthesized a series of multivalent N-acetylglucosamine (GlcNAc) derivatives and studied

their interaction with the plant lectin wheat germ agglutinin (WGA) by an enzyme-linked lectin assay (ELLA)

and X-ray crystallography. The solution conformation of one ligand was determined by NMR spectroscopy.

Employing a GlcNAc carbamate motif with R-conﬁguration and by systematic variation of the spacer length,

we were able to identify divalent ligands with unprecedented high WGA binding potency. The best divalent

ligand has an IC₅₀value of 9.8 µM (ELLA) corresponding to a relative potency of 2350 (1170 on a valency-

corrected basis, i.e., per mol sugar contained) compared to free GlcNAc. X-ray crystallography of the complex

of WGA and the second best, closely related divalent ligand explains this activity. Four divalent molecules

simultaneously bind to WGA with each ligand bridging adjacent binding sites. This shows for the ﬁrst time

that all eight sugar binding sites of the WGA dimer are simultaneously functional. We also report a tetravalent

neoglycopeptide with an IC₅₀value of 0.9 µM being 25 500 times higher than that of GlcNAc (6400 times

per contained sugar) and the X-ray structure analysis of its complex with glutaraldehyde-cross-linked WGA.

Comparison of the crystal structure and the solution NMR structure of the neoglycopeptide as well as

results from the ELLA suggest that the conformation of the glycopeptide in solution is already preorganized

in a way supporting multivalent binding to the protein. Our ﬁndings show that bridging adjacent protein

binding sites by multivalent ligands is a valid strategy to ﬁnd high-afﬁnity protein ligands and that even

subtle changes of the linker structure can have a signiﬁcant impact on the binding afﬁnity.

sites on DNA.⁶Multivalent interactions are characterized by

the simultaneous binding of multiple epitopes of one entity

Introduction

Multivalent interactions are frequently observed in biological

systems where they govern important recognition processes.¹

Carefully studied examples include the adhesion of inﬂuenza

viruses to bronchial epithelial cells,²adhesion of uropatho-

genic Escherichia coli strains to urethral endothelial cells,³

neutrophil-endothelium interactions during the inﬂammatory

process,⁴the binding of cholera toxin to gangliosides GM1 on

cell surfaces,⁵and the binding of transcription factors to multiple

(molecule, surface) with several binding sites of another and

are responsible for the enhancement of weak interactions both

in solution and at interfaces. Furthermore, they result in binding

kinetics different from monovalent interactions⁷and provide a

means for the ﬁne-tuning of such processes.

Many examples of multivalent binding involve the interaction

between carbohydrates and proteins.⁸In these cases, increased

binding afﬁnity due to multivalency has been termed the cluster

glycoside effect.^8e,9Multidentate carbohydrate ligands are of

signiﬁcant medical interest in the diagnosis and inhibition of

^†Department of Biology, Universita¨t Konstanz.

^‡Department of Chemistry, Universita¨t Konstanz.

^§Current address: Merck KGaA, Frankfurter Str. 250, 64293 Darmstadt,

Germany.

(6) von Hippel, P. H. Annu. ReV. Biophys. Biomol. Struct. 2007, 36, 79–

105.

(1) (a) Mammen, M.; Choi, S.-K.; Whitesides, G. M. Angew. Chem., Int.

Ed. 1998, 37, 2755–2794. (b) Huskens, J. Curr. Opin. Chem. Biol.

2006, 10, 537–543. (c) Whitty, A. Nat. Chem. Biol. 2008, 4, 435–

439.

(7) Rao, J.; Lahiri, J.; Weis, R. M.; Whitesides, G. M. J. Am. Chem. Soc.

2000, 122, 2698–2710.

(8) (a) Imberty, A.; Varrot, A. Curr. Opin. Struct. Biol. 2008, 18, 567–

576. (b) Dam, T. K.; Brewer, C. F. In ComprehensiVe Glycoscience;

Kamerling, J. P., Ed.; Elsevier: Oxford, 2007; pp 397-452. (c) Collins,

B. E.; Paulson, J. C. Curr. Opin. Chem. Biol. 2004, 8, 617–625. (d)

Sacchettini, J. C.; Baum, L. G.; Brewer, C. F. Biochemistry 2001, 40,

3009–3015. (e) Lee, Y. C.; Lee, R. T. Acc. Chem. Res. 1995, 28, 321–

327. (f) Drickamer, K.; Taylor, M. E. Annu. ReV. Cell Biol. 1993, 9,

237–264.

(2) Lees, W. J.; Spaltenstein, A.; Kingery-Wood, J. E.; Whitesides, G. M.

J. Med. Chem. 1994, 37, 3419–3433.

(3) Connell, I.; Agace, W.; Klemm, P.; Schembri, M.; Ma˘rild, S.;

Svanborg, C. Proc. Natl. Acad. Sci. U.S.A. 1996, 93, 9827–9832.

(4) (a) Varki, A. J. Clin. InVest. 1997, 99, 158–162. (b) Lasky, L. A.

Annu. ReV. Biochem. 1995, 64, 113–139.

(5) Fan, E.; Merritt, E. A.; Verlinde, C. L. M. J.; Hol, W. G. J. Curr.

Opin. Struct. Biol. 2000, 10, 680–686.

(9) Lundquist, J. J.; Toone, E. J. Chem. ReV. 2002, 102, 555–578.

9

8704 J. AM. CHEM. SOC. 2010, 132, 8704–8719

10.1021/ja101646k  2010 American Chemical Society

Multivalent Binding to Wheat Germ Agglutinin

A R T I C L E S

such processes or as effectors of signal transduction pathways.^8c,10

The enhancement of binding afﬁnity toward target proteins as

a result of ligand clustering can be remarkable and has been

determined by different types of binding assays.⁹In many cases,

however, the molecular principles of how binding enhancement

is achieved are not well understood. Several mechanisms can

account for the observed binding enhancements.^10i,11Generally,

the highest contribution is attributed to the chelate effect, that

is, the spanning of adjacent binding sites by the multivalent

ligand. It is obvious that the nature of the spacer between the

binding epitopes, especially its conformation and ﬂexibility,

strongly inﬂuences the ligand’s chelate formation capability.

However, signiﬁcant binding enhancements are also observed

in cases where the spacer is too short to allow spanning of

adjacent binding sites. In these cases, other mechanisms are

assumed to be operative. One of them is statistical rebinding,

that is, the rebinding of epitopes before complete dissociation

of the complex due to high local concentration of epitopes.

Often, cross-linking of proteins by the multivalent ligand leading

to lattice formation is observed.¹²While this process has been

studied in detail, the mechanisms of the often associated

enhanced binding afﬁnities are less well understood.

closely related AB₅bacterial toxin.¹⁸Divalent inhibitors of the

cholera toxin B pentamer were shown to have high potency without

the ability to span two binding sites.¹⁹

Recently, we introduced a diversity-oriented approach for the

combinatorial synthesis and screening of multivalent lectin ligands

with varying scaffold architectures.²⁰In this approach, termed

spatial screening of multivalent lectin ligands, cyclopeptides serve

as scaffolds for the presentation of carbohydrate epitopes. From

screening of a one-bead one-compound library of almost 20 000

cyclic neoglycopeptides, we were able to identify several tetra- to

hexavalent wheat germ agglutinin (WGA) ligands with strongly

increased binding potencies compared to monovalent N-acetylglu-

cosamine (GlcNAc).^20bWe concluded that it is the spatial presenta-

tion of the GlcNAc residues on the cyclic peptide scaffolds that is

responsible for their high afﬁnity.

WGA is a plant lectin that that is enriched in the seeds of

Triticum Vulgaris and exists in three closely related major

isoforms, WGA1, WGA2, and WGA3. It is speciﬁc for terminal

N-acetylneuraminic acid and GlcNAc and has been shown to

inhibit fungal growth through interaction with fungal cell-wall

components²¹and to agglutinate transformed cells in Vitro.²²

WGA forms a 36-kDa stabile homodimer with a twofold

symmetry axis.²³Each polypeptide chain forms four hevein

domains (43 residues each), A-D. The carbohydrate binding

sites were already analyzed by cocrystallization of the protein

with sialyl lactose (Neu5Ac-R(2,3)-Gal-ꢀ(1,4)-Glc)²⁴and with a

sialoglycopeptide.²⁵Furthermore, carbohydrate-protein interaction

was examined by soaking of glutaraldehyde-cross-linked WGA

crystals with N,N′-diacetyl chitobiose (GlcNAc-ꢀ(1,4)-GlcNAc),^23a,26

GlcNAc-ꢀ(1,6)-Gal,^26band GlcNAc-ꢀ(1,6)-Gal-ꢀ(1,4)-Glc.^26bCom-

bined evidence from these investigations revealed eight functional

sugar binding sites per WGA dimer (four unique sites due to the

twofold symmetry axis). However, simultaneous occupancy of all

eight sites has not been observed in a single crystal structure. A

binding site for GlcNAc or ꢀ-(1,4)-linked GlcNAc oligomers²⁷is

formed by each hevein domain via a cluster of three conserved

aromatic residues of which the second is stacking to the sugar ring.

Binding is complemented by polar residues from an adjacent

domain of the other chain which provides hydrogen bonds. The

sites can be adequately labeled as pairs of capitals (each followed

by the number of the polypeptide chain it belongs to) indicating

the hevein subunit which provides the aromatic residues and the

polar residues, respectively. In D, the polar residues are absent, so

Structural information on multivalent carbohydrate-protein

interactions with atomic resolution is rare. Cross-linking of lectins

by multivalent ligands has been observed for galectin-1¹³and

concanavalin A dimers,¹⁴monomers of the carbohydrate recogni-

tion domain of DC-SIGN,¹⁵and BC2L-A dimers.¹⁶The only

examples of lectin structures with bound ligands that span several

binding sites of a single (oligomeric) lectin stem from the AB₅

family of bacterial toxins. Kitov et al. designed a decavalent ligand,

named STARFISH, for Shiga-like toxins with subnanomolar

binding potency.¹⁷Deviant from the original design, crystal

structure analysis revealed that the ligand is sandwiched between

two toxin B pentamer molecules. The pairs of trisaccharides at

the tips of each of the ﬁve spacer arms connect pairs of opposite

B subunits of the two toxin pentamers. Similar binding modes were

observed for penta- and decavalent ligands for the cholera toxin, a

(10) (a) Neoglycoconjugates: Preparation and Applications; Lee, Y. C.;

Lee, R. T., Eds.; Academic Press: San Diego, CA, 1994; (b) Roy, R.

Top. Curr. Chem. 1997, 187, 241–274. (c) Houseman, B. T.; Mrksich,

M. Top. Curr. Chem. 2002, 218, 1–44. (d) Lindhorst, T. K. Top. Curr.

Chem. 2002, 218, 201–235. (e) Ortiz Mellet, C.; Defaye, J.; Garc´ıa

Ferna´ndez, J. M. Chem.sEur. J. 2002, 8, 1982–1990. (f) Turnbull,

W. B.; Stoddart, J. F. ReV. Mol. Biotechnol. 2002, 90, 231–255. (g)

Choi, S.-K. Synthetic MultiValent Molecules. Concepts and Biomedical

Applications; John Wiley & Sons: Hoboken, NJ, 2004; (h) Wittmann,

V. In Highlights in Bioorganic Chemistry: Methods and Applications;

Schmuck, C., Wennemers, H., Eds.; Wiley-VCH: Weinheim, 2004;

pp 203-213. (i) Kiessling, L. L.; Gestwicki, J. E.; Strong, L. E. Angew.

Chem., Int. Ed. 2006, 45, 2348–2368.

(18) (a) Merritt, E. A.; Zhang, Z.; Pickens, J. C.; Ahn, M.; Hol, W. G. J.;

Fan, E. J. Am. Chem. Soc. 2002, 124, 8818–8824. (b) Zhang, Z.;

Merritt, E. A.; Ahn, M.; Roach, C.; Hou, Z.; Verlinde, C. L. M. J.;

Hol, W. G. J.; Fan, E. J. Am. Chem. Soc. 2002, 124, 12991–12998.

(19) Pickens, J. C.; Mitchell, D. D.; Liu, J.; Tan, X.; Zhang, Z.; Verlinde,

C. L. M. J.; Hol, W. G. J.; Fan, E. Chem. Biol. 2004, 11, 1205–1215.

(20) (a) Wittmann, V.; Seeberger, S. Angew. Chem., Int. Ed. 2000, 39,

4348–4352. (b) Wittmann, V.; Seeberger, S. Angew. Chem., Int. Ed.

2004, 43, 900–903.

(11) (a) Dam, T. K.; Brewer, C. F. Biochemistry 2008, 47, 8470–8476. (b)

Pieters, R. J. Org. Biomol. Chem. 2009, 7, 2013–2025.

(12) Brewer, C. F.; Miceli, M. C.; Baum, L. G. Curr. Opin. Struct. Biol.

2002, 12, 616–623.

(21) (a) Mirelman, D.; Galun, E.; Sharon, N.; Lotan, R. Nature 1975, 256,

414–416. (b) Laroy, K.; Weeks, G. J. Cell Sci. 1983, 59, 203–212.

(c) Asensio, J. L.; Canada, F. J.; Siebert, H. C.; Laynez, J.; Poveda,

A.; Nieto, P. M.; Soedjanaamadja, U. M.; Gabius, H. J.; Jimenez-

Barbero, J. Chem. Biol. 2000, 7, 529–543.

(13) Bourne, Y.; Bolgiano, B.; Liao, D.-I.; Strecker, G.; Cantau, P.;

Herzberg, O.; Feizi, T.; Cambillau, C. Nat. Struct. Biol. 1994, 1, 863–

870.

(22) (a) Aub, J. C.; Sanford, B. H.; Cote, M. N. Proc. Natl. Acad. Sci.

U.S.A. 1965, 54, 396–399. (b) Aub, J. C.; Sanford, B. H.; Wang, L. H.

Proc. Natl. Acad. Sci. U.S.A. 1965, 54, 400–402.

(14) Dimick, S. M.; Powell, S. C.; McMahon, S. A.; Moothoo, D. N.;

Naismith, J. H.; Toone, E. J. J. Am. Chem. Soc. 1999, 121, 10286–

10296.

(23) (a) Wright, C. S. J. Mol. Biol. 1980, 141, 267–291. (b) Wright, C. S.

J. Mol. Biol. 1989, 209, 475–487. (c) Harata, K.; Nagahora, H.; Jigami,

Y. Acta Crystallogr., Sect. D: Biol. Crystallogr. 1995, 51, 1013–1019.

(24) Wright, C. S. J. Mol. Biol. 1990, 215, 635–651.

(15) Feinberg, H.; Mitchell, D. A.; Drickamer, K.; Weis, W. I. Science

2001, 294, 2163–2166.

(16) Lameignere, E.; Shiao, T. C.; Roy, R.; Wimmerova, M.; Dubreuil,

F.; Varrot, A.; Imberty, A. Glycobiology 2010, 20, 87–98.

(17) Kitov, P. I.; Sadowska, J. M.; Mulvery, G.; Armstrong, G. D.; Ling,

H.; Pannu, N. S.; Read, R. J.; Bundle, D. R. Nature 2000, 403, 669–

672.

(25) Wright, C. S.; Jaeger, J. J. Mol. Biol. 1993, 232, 620–638.

(26) (a) Wright, C. S. J. Mol. Biol. 1984, 178, 91–104. (b) Muraki, M.;

Ishimura, M.; Harata, K. Biochim. Biophys. Acta 2002, 1569, 10–20.

(27) Wright, C. S.; Kellogg, G. E. Protein Sci. 1996, 5, 1466–1476.

9

J. AM. CHEM. SOC. VOL. 132, NO. 25, 2010 8705

A R T I C L E S

Schwefel et al.

Chart 1. Tetravalent WGA-Binding Neoglycopeptides

that the sites are A1, B1C2, C1B2, D1A2, A2, B2C1, C2B1, D2A1.

The binding sites involving A and D exhibit lower afﬁnity as those

formed from B and C.²⁷Consequently, the sites involving the B

and C domains have been termed “primary” binding sites and the

ones involving the A and D domains “secondary” binding sites.

The secondary binding sites are generally assumed to bind

carbohydrates with an afﬁnity too weak to be detected in solution.

Out of the WGA ligands identiﬁed by our spatial screening

procedure,^20btetravalent neoglycopeptide 1 (Chart 1) showed

an unprecedentedly strong increase in WGA binding potency

by a factor of 1440 (360 per sugar residue) compared to

monovalent GlcNAc as determined by an enzyme-linked lectin

assay (ELLA).²⁸In this assay, the concentration of ligand

leading to an inhibition of 50% of the binding of horseradish

peroxidase-labeled WGA to GlcNAc residues covalently im-

mobilized to microtiter plates (IC₅₀value) is determined.

Assuming that the spatial presentation of the GlcNAc residues

on the cyclic peptide scaffold is responsible for its high afﬁnity,

variation of the length and ﬂexibility of the spacer between the

GlcNAc residues and the peptide backbone of 1 is expected to

strongly affect the IC₅₀value of 16 µM of the glycocluster.

In this report, we describe the synthesis of cyclic neoglycopeptide

2 (Chart 1) with a spacer length reduced by ﬁve bonds. In contrast

to 1, the GlcNAc residues in 2 are R-glycosidically linked to the

peptide scaffold. The new ligand 2 exhibits an improved IC₅₀value

of 0.9 µM that is 25 500-fold lower than that of GlcNAc (or 6400-

fold on a valency-corrected basis). To reveal the structural basis

of the enhanced WGA binding afﬁnity of 2, we prepared a series

of mono- to trivalent GlcNAc derivatives containing the glycosyl

carbamate motive found in 2 separated by different spacers. WGA

binding afﬁnities were determined by ELLA. Systematic variation

of spacing identiﬁed the optimal distance between GlcNAc residues

to span adjacent binding sites in WGA. The best divalent ligand

has an IC₅₀value of 9.8 µM which is even lower than that of

tetravalent glycopeptide 1. We also report a high-resolution crystal

structure of the complex between a divalent ligand and the lectin.

In this structure, four pairs of adjacent binding sites are bridged

by four molecules of the divalent ligand. This is the ﬁrst structure

of WGA in which all eight binding sites are occupied. In addition,

it is the ﬁrst example of a crystal structure of a lectin in complex

with several chelating multivalent carbohydrate ligands. Further-

more, we report the crystal structure of the complex of tetravalent

neoglycopeptide 2 and WGA obtained from a soaking experiment.

Comparison of the peptide conformation seen in the complex with

the NMR-derived solution structure of 2 suggests an explanation

for the high afﬁnity of the glycopeptide.

Results and Discussion

Synthesis of Ligands. Glycosyl p-nitrophenyl carbonates are

suitable precursors for bioconjugation of carbohydrates.²⁹Their

reaction with amines is fast, produces glycosyl carbamates in

high yields, and proceeds under retention of conﬁguration at

the anomeric center. Thus, glycosyl p-nitrophenyl carbonates

(28) Maierhofer, C.; Rohmer, K.; Wittmann, V. Bioorg. Med. Chem. 2007,

15, 7661–7676.

9

8706 J. AM. CHEM. SOC. VOL. 132, NO. 25, 2010

Multivalent Binding to Wheat Germ Agglutinin

A R T I C L E S

Scheme 1. Synthesis of Glycosyl Carbonate 5

Scheme 2. Synthesis of Neoglycopeptide 2

are excellent precursors for the stereoselective synthesis of

O-glycosyl carbamates, provided they are available as pure

isomers. Recently, we could show that both R- and ꢀ-lactosyl

p-nitrophenyl carbonates are accessible as pure isomers by

proper choice of reaction conditions.^29eWe now report condi-

tions for the stereoselective synthesis of GlcNAc-derived

p-nitrophenyl carbonate 5 in high yield (Scheme 1). Peracety-

lated GlcNAc 3 was selectively deprotected at the anomeric

center using the method of Kova´cˇ.³⁰Resulting 4 exists according

1

to H NMR analysis (CDCl₃) as mixture of anomers (R/ꢀ )

9:1). Treatment with p-nitrophenyl chloroformate and triethyl-

amine in dichloromethane exclusively led to R-glycosyl carbon-

ate 5 in a yield of 92%. Since 5 slowly decomposes during

column chromatography on silica gel, it was essential to use a

short column and avoid prolonged contact of 5 with the silica

gel.

To probe the effect of the length and ﬂexibility of the spacer

between the GlcNAc residues and the peptide backbone of 1

on the WGA binding afﬁnity, we synthesized neoglycopeptide

2. The carbohydrate-peptide spacer of 2 is shortened by ﬁve

bonds and also is less ﬂexible due to the missing butene-diol

moiety. Furthermore, the GlcNAc residues are R-glycosidically

linked to the peptide scaffold leading to a different spatial

orientation of the sugars. It is known that WGA binds both R-

and ꢀ-glycosides of GlcNAc with similar afﬁnity.³¹Synthesis

of 2 started from TentaGel resin 6 loaded with ꢀ-alanine via

the Sieber linker³²(Scheme 2). Assembly of linear peptide 7

followed the Fmoc strategy;³³solely in the last coupling step

an N^R-Boc-protected amino acid (Boc-Lys(Aloc)-OH) was

applied. Removal of the Aloc group and cleavage of the allyl

ester were achieved by treatment with [Pd(PPh₃)₄] and borane

dimethylamine complex.³⁴Subsequent addition of a mixture of

2-(1H-benzotriazol-1-yl)-1,1,3,3-tetramethyluronium hexaﬂuo-

rophosphate (HBTU),³⁵1-hydroxybenzotriazole (HOBt), and

Hu¨nig’s base (EtNi-Pr₂) gave cyclic peptide 8.

We found that usage of resin with a low loading (0.2 mmol

g^-1) is crucial during the cyclization step. Only in this way, it

was possible to obtain high yields of cyclic peptide as

determined by a test cleavage from a small sample of resin 8.

With higher resin loadings, we were not able to cleave any

material from 8 by treatment with 1% TFA. A plausible

explanation for this observation is the “intermolecular” reaction

of peptide molecules during the cyclization step leading to a

solid phase-bound peptide polymer. At lower resin loadings,

cyclization of single linear peptides is preferred, an observation

which is referred to as the pseudodilution phenomenon.³⁶

Cleavage of the 1-(4,4-dimethyl-2,6-dioxocyclohexylidene)isov-

aleryl (Ddv) groups³⁷that were used to mask the γ-NH₂groups

of D-diaminobutyric acid (D-Dab) was effected by batch-wise

treatment with a 4% solution of hydrazine hydrate which turned

out^20ato be advantageous over the literature-recommended³⁷

2% solution. Subsequently, the GlcNAc residues were attached

by addition of glycosyl carbonate 5 (3 equiv. per free NH₂

group) in the presence of Hu¨nig’s base until Kaiser³⁸and TNBS

test³⁹indicated the absence of free amino groups. Finally, the

peptide was cleaved from the resin with 1% triﬂuoroacetic acid

(TFA) and deacetylated to give 2. It is important to carry out

the deacetylation after cleavage from the resin because the

glycosidic bond of the deacetylated sugars is not completely

(29) (a) Madec-Lougerstay, R.; Florent, J.-C.; Monneret, C. J. Chem. Soc.,

Perkin Trans. 1 1999, 1369–1376. (b) Azoulay, M.; Escriou, V.;

Florent, J.-C.; Monneret, C. J. Carbohydr. Chem. 2001, 20, 841–853.

(c) Cherif, S.; Leach, M. R.; Williams, D. B.; Monneret, C. Bioorg.

Med. Chem. Lett. 2002, 12, 1237–1240. (d) Cai, T. B.; Lu, D.; Tang,

X.; Zhang, Y.; Landerholm, M.; Wang, P. G. J. Org. Chem. 2005,

70, 3518–3524. (e) Andre´, S.; Specker, D.; Bovin, N. V.; Lensch,

M.; Kaltner, H.; Gabius, H.-J.; Wittmann, V. Bioconjugate Chem.

2009, 20, 1716–1728.

(30) Zhang, J.; Kova´cˇ, P. J. Carbohydr. Chem. 1999, 18, 461–469.

(31) Allen, A. K.; Neuberger, A.; Sharon, N. Biochem. J. 1973, 131, 155–

162.

(36) (a) Mazur, S.; Jayalekshmy, P. J. Am. Chem. Soc. 1979, 101, 677–

683. (b) Albericio, F.; Hammer, R. P.; Garcia-Echeverria, C.; A.,

M. M.; Chang, J. L.; Munson, M. C.; Pons, M.; Giralt, E.; Barany, G.

Int. J. Pept. Protein Res. 1991, 37, 402–413.

(32) Sieber, P. Tetrahedron Lett. 1987, 28, 2107–2110.

(33) Wellings, D. A.; Atherton, E. Methods Enzymol. 1997, 289, 44–67.

(34) (a) Gomez-Martinez, P.; Dessolin, M.; Guibe´, F.; Albericio, F. J. Chem.

Soc., Perkin Trans. 1 1999, 2871–2874. (b) Jarikote, D. V.; Ko¨hler,

O.; Socher, E.; Seitz, O. Eur. J. Org. Chem. 2005, 318, 7–3195.

(35) Knorr, R.; Trzeciak, A.; Bannwarth, W.; Gillessen, D. Tetrahedron

Lett. 1989, 30, 1927–1930.

(37) Chhabra, S. R.; Hothi, B.; Evans, D. J.; White, P. D.; Bycroft, B. W.;

Chan, W. C. Tetrahedron Lett. 1998, 39, 1603–1606.

(38) Kaiser, E.; Colescott, R. L.; Bossinger, C. D.; Cook, P. I. Anal.

Biochem. 1970, 34, 595–598.

(39) Hancock, W. S.; Battersby, J. E. Anal. Biochem. 1976, 71, 260–264.

9

J. AM. CHEM. SOC. VOL. 132, NO. 25, 2010 8707

A R T I C L E S

Schwefel et al.

Scheme 3. Conjugation of Glycosyl Carbonate 5 to Amines 10-14

Scheme 4. Synthesis of Glycoclusters 20-24^a

stable under the acidic conditions used for cleavage of the

peptide from the resin.^20a,40

A series of mono- to trivalent GlcNAc derivatives containing

the R-glycosyl carbamate motive found in 2 was obtained by

conjugation of carbonate 5 to commercially available amines

10-14 (Scheme 3). Couplings were carried out in dichlo-

romethane in the presence of Hu¨nig’s base to give carbamates

15-19 in very good yields. Deprotection by treatment with

EtNMe₂in MeOH or under Zemple´n conditions⁴¹yielded

glycoclusters 20-24 (Scheme 4).

^a(a) EtNMe₂, MeOH; (b) NaOMe, MeOH.

Scheme 5. Synthesis of Divalent Neoglycopeptide 28

As detailed below, binding of neoglycopeptide 2 to WGA

occurs by spanning adjacent binding sites through the two

GlcNAc residues bound to D-Dab residues in position 4 and 5.

Therefore, we also prepared the divalent glycopeptide 28

representing this substructure of 2 (Scheme 5). Starting from

Fmoc-protected Rink Amide polystyrene resin 25, two D-Dab

residues were coupled by the Fmoc strategy followed by

replacement of the N-terminal Fmoc group with an acetyl group.

Ddv removal, sugar conjugation, and cleavage from the resin

with concentrated TFA gave divalent glycopeptide 27 which

was deacetylated to yield 28.

Binding Assays. WGA binding potencies of the synthetic

ligands were measured in terms of IC₅₀values for the inhibition

of binding of horseradish peroxidase-labeled WGA to a GlcNAc

derivative covalently immobilized to microtiter plates using an

ELLA developed earlier.²⁸As listed in Table 1, monovalent

ligand 20 has an afﬁnity similar to that of ꢀ-allyl glycoside 29

within the margin of error of the ELLA. Therefore, we conclude

that it is not the anomeric conﬁguration or the glycosyl

carbamate structure that is responsible for the strongly increased

binding afﬁnity of glycocluster 2 compared to 1 but rather the

altered spatial presentation of the sugars on the peptide scaffold

and/or the reduced ﬂexibility of 2.

The inﬂuence of sugar spacing on the WGA afﬁnity is clearly

seen by comparison of divalent ligands 21-23. Whereas divalent

(40) Kunz, H.; Unverzagt, C. Angew. Chem., Int. Ed. Engl. 1988, 27, 1697–

1699.

ligand 21 showed only 31-fold afﬁnity increase over GlcNAc

(11-fold increase over corresponding monovalent 20), incorpo-

(41) Zemple´n, G.; Pacsu, E. Ber. Dtsch. Chem. Ges. 1929, 62, 1613–1614.

9

8708 J. AM. CHEM. SOC. VOL. 132, NO. 25, 2010

Multivalent Binding to Wheat Germ Agglutinin

A R T I C L E S

Table 1. IC₅₀Values of Synthetic Ligands for Inhibition of Binding of Horseradish Peroxidase-Labeled WGA to GlcNAc Residues Covalently

Immobilized to Microtiter Plates from Dose-Response Curves Shown in Figures S1-S3 Determined by an ELLA²⁸

^aRelative potencies are referenced to GlcNAc.

Crystal Structure of WGA1 in Complex with Propyl

Carbamate 20. We ﬁrst determined the structure of WGA1 in

complex with propyl carbamate 20 by cocrystallization

(WGA1∼20, PDB ID: 2UWG). This structure served as a

reference for structures of complexes of WGA with multivalent

ligands by allowing to assess whether the scaffolds of the

multivalent ligands have any unfavorable impact on the sugar

binding mode. The crystals belonged to space group P3₂21 and

diffracted to 1.6 Å resolution (Table 2). The structure was

reﬁned to an R (R_free) factor of 19.6% (23.2%). One WGA1

homodimer was found in the asymmetric unit with a conforma-

tion identical to unliganded WGA1 (PDB ID: 7WGA²⁴). All

high afﬁnity sites B1C2, C1B2, B2C1, and C2B1 of the WGA1

dimer were found occupied with one molecule of 20 (Figure

ration of 7 additional bonds in the spacer results in 22 with a

remarkable IC₅₀value of 9.8 µM corresponding to a relative

potency of 2350 over GlcNAc (816 over 20). Further, increase

of the spacer length by only one additional bond leads to a drop

of relative potency to 400 over GlcNAc (140 over 20) for ligand

23 which is, however, still unprecedentedly high for a divalent

WGA ligand.^11bThe potency of 23 is even better than that of

trivalent 24 which has the same spacer length between two

GlcNAc moieties as 21. Possibly, this linker length is not

sufﬁcient to span adjacent binding sites of the lectin. To gain

more insight into the structural details of multivalent binding

to WGA, we determined the crystal structure of several

WGA-ligand complexes. All structures were solved by mo-

lecular replacement.

9

J. AM. CHEM. SOC. VOL. 132, NO. 25, 2010 8709

A R T I C L E S

Schwefel et al.

Table 2. Crystallographic Data

WGA1∼20

WGA3∼23

WGA1∼2

PDB ID

2UWG

2X52

2X3T

X-ray source

X-ray wavelength [Å]

Space group

Cell dimensions

a [Å]

SLS X06SA

1

P3₂21

ESRF ID23-1

1

R3

ESRF ID23-1

1

P2₁

63.2

101.4

44.6

b [Å]

63.2

101.4

93.2

c [Å]

153.4

145.0

92.5

R [deg]

90.0

120.0

90.0

120.0

20-1.70 (1.75-1.70)

315384 (12253)

60413 (4459)

11.54 (2.06)

98.9 (87.1)

10.2 (62.3)

26

90.0

98.3

90.0

ꢀ [deg]

γ [deg]

Resolution [Å]

Number of measured reﬂections

Number of unique reﬂections

I/σ(I)

44.59-1.60 (1.65-1.60)^a

438743 (30984)

46032 (3729)

17.38 (3.52)

96.1 (89.9)

6.6 (62.0)

33

50-2.75 (2.91-2.75)

73664 (11828)

19531 (3124)

12.10 (3.45)

99.5 (99.2)

10.3 (49.2)

54

Completeness [%]

Rmeas [%]^b

Wilson B-factor [Å²]

Twinning

Twin law

Twin fraction

-h-k,k,-l

0.44

Reﬁnement

Number of reﬂections used

Number of protein atoms

Number of ligand atoms

Number of water molecules

Final R-factor [%]

43465

2307

87

245

19.6

23.2

57320

2330

182

178

15.8

17.1

19531

4521

261

38

22.9

28.5

Final free R-factor [%]

Ramachandran plot

Favored [%]

Outliers [%]

97.3

0.0

97.9

0.0

90.3

0.4

Rms deviation

Bond lengths [Å]

Bond angles [deg]

0.012

1.46

0.006

1.04

0.002

0.72

Average B-factor [Å²]

Protein

Ligand

31.4

32.0

32.4

23.7

25.8

28.4

54.7

66.6

43.7

Water

^aAll values in parentheses apply for the highest resolution shell. ^bAccording to Diederichs and Karplus⁴³

1A and Figure S4A). Because of a crystal contact, site A1 could

not accommodate a ligand. At the low afﬁnity sites D1A2, A2,

and D2A1, blobs of F_o- F_cdifference electron density were

found after molecular replacement, too big to be water.

However, fully occupied molecules of 20 could not be placed

there unambiguously. Nevertheless, in cocrystallization experi-

ments with higher concentrations of 20 (50 mM), all sites except

A1 were found occupied (data not shown). Comparison with

the structures of the complexes of WGA3 with GlcNAc-ꢀ(1,4)-

GlcNAc^26band WGA1 with GlcNAc⁴²shows that binding of

20 to WGA1 is almost identical to GlcNAc binding (Figure

1B).

Crystal Structure of WGA3 in Complex with Divalent

Ligand 23. WGA3 in complex with divalent 23 (WGA3∼23,

PDB ID: 2X52) crystallized in space group R3. Crystals

diffracted to 1.7 Å (Table 2). The single crystal used for data

collection was merohedrally twinned, with a twinning fraction

of 0.44. The model could be reﬁned to a ﬁnal R (R_free) factor of

15.8% (17.1%). One WGA3 dimer was found in the asymmetric

unit in identical conformation as in WGA1∼20, except for the

minor sequence differences between the WGA isoforms 1 and

3.

All eight proposed carbohydrate binding sites of the WGA3

dimer²⁷were found occupied by the GlcNAc-carbamate moieties

of four molecules of 23 with each divalent ligand spanning pairs

of adjacent binding sites (Figure 2 and Figure S4B). This shows

for the ﬁrst time that all sugar binding sites of the WGA dimer

are simultaneously functional. The arrangement of the GlcNAc

moieties in the sites is the same as in the WGA1∼20 structure.

Two divalent ligands bind to the high-afﬁnity site pairs

B1C2-C2B1 and B2C1-C1B2, respectively. The conformation

of the linker region of these ligands is well-deﬁned but differs

for the two ligands (Figure 3). The differences are likely due to

participation of the former ligand in a crystal contact (Figure

3A). Residues Arg-139 and Gln-6 from polypeptide chain 1 of

a crystallographically related WGA dimer form hydrogen bonds

to carbamate oxygens at both ends of the linker region. Being

constrained in this way, the linker region adopts a distinct

conformation, which allows interaction with Trp-107 of chain

2 via water coordination. The linker region of the divalent ligand

at site pair B2C1-C1B2 (Figure 3B) does not interact with

(42) Schwefel, D.; Wittmann, V.; Diederichs, K.; Welte, W. PDB ID: 2UVO

(WGA1 in complex with GlcNAc), to be published.

(43) Diederichs, K.; Karplus, P. A. Nat. Struct. Biol. 1997, 4, 269–275.

9

8710 J. AM. CHEM. SOC. VOL. 132, NO. 25, 2010

Multivalent Binding to Wheat Germ Agglutinin

A R T I C L E S

Figure 4. Portions of cyclic peptide 2 bound to WGA1 (PDB ID: 2X3T),

which are visible in the crystal structure. The protein is shown as surface

representation. WGA chain 1 is colored green, chain 2 is cyan. Ligands

are shown as stick model with carbons colored white, nitrogens colored

blue, and oxygens in red.

Figure 1. Crystal structure of WGA1 in complex with propyl carbamate

20 (PDB ID: 2UWG). (A) The protein is shown as surface representation.

WGA1 chain 1 is colored green, chain 2 cyan. Compound 20 is shown in

stick representation. Carbon atoms are colored white, oxygens red, and

nitrogens blue. (B) Comparison of the binding mode of compound 20 to

the GlcNAc binding mode (PDB ID: 2UVO⁴²) at site C1B2. 20 and

interacting WGA chain 1 residues are colored gray; interacting WGA chain

2 residues are dark blue. GlcNAc and interacting WGA chain 1 residues

are colored green; interacting WGA chain 2 residues are light blue. The

digits behind the residue numbers designate the WGA chains 1 and 2.

Hydrogen bonds are indicated as magenta lines.

density. In this case, the middle part of the linker could not be

modeled (Figure 2). This indicates higher static or dynamic

disorder of these linker regions, possibly due to the smaller

binding afﬁnity of the GlcNAc-carbamate moieties compared

to ligands at high-afﬁnity binding sites. Nevertheless, intermo-

lecular bridging to symmetry-related bindings sites of adjacent

WGA dimers can be excluded because of the arrangement of

binding sites.

Crystal Structure of WGA1 in Complex with Tetravalent

Cyclic Neoglycopeptide 2 (WGA1∼2, PDB ID: 2X3T). Glut-

araldehyde-cross-linked WGA1 crystals were soaked with cyclic

tetravalent glycopeptide 2. The crystals belonged to space group

P2₁and diffracted to 2.75 Å. The structure was reﬁned to a

ﬁnal R (R_free) factor of 22.9% (28.5%) (Table 2). Two WGA1

dimers were found in the asymmetric unit. The polypeptide

chains of the ﬁrst dimer were designated as 1 and 2, the

polypeptides of the second as 3 and 4. Eleven out of 16 possible

sugar binding sites of the two dimers were found occupied. The

remaining ﬁve sites were obstructed by crystal contacts. At the

occupied sites clear electron density extended beyond the

anomeric oxygen of the GlcNAc moiety and the complete

GlcNAc-carbamate moiety could be modeled. Moreover, at two

pairsoftheoccupiedsites,namely,B1C2-C2B1andB3C4-C4B3,

the complete glycosylated D-Dab residues 4 and 5 of two cyclic

neoglycopeptides 2 are visible (Figure 4 and Figure S4C). These

glycosylated dipeptide subunits of 2 bridge the site pairs in a

similar way as does the divalent ligand in complex WGA3∼23

(Figure 5). The conformation of both of these dipeptide subunits

is identical within the limits of experimental data. Further

Figure 2. Crystal structure of WGA3∼23 (PDB ID: 2X52). The protein

is shown as surface representation. WGA chain 1 is colored green, chain 2

is cyan. The divalent ligand is shown as stick model with carbons colored

white, nitrogens colored blue, and oxygens in red.

protein residues at all. Its ﬁxation via carbohydrate-WGA

interaction is apparently sufﬁcient to deﬁne its conformation.

This conformation should, thus, be relevant for protein-ligand

interaction in solution. Linker regions from ligands bridging

low afﬁnity binding sites (Ax, DxAy) show only weak electron

Figure 3. Detailed view of divalent ligand 23 bound at (A) binding site pair B1C2-C2B1 and (B) B2C1-C1B2 of WGA3. Compound 23 is shown in

ball-and-stick representation with carbon atoms colored white, nitrogens blue, and oxygens red. Interacting protein side chains are shown in stick representation.

Residues from chain 1 are colored green, those from chain 2 are cyan. Hydrogen bonds are indicated as magenta lines. The digits behind the residue numbers

designate the WGA chains 1 and 2, respectively, and “sym” indicates residues from a crystallographically related WGA dimer.

9

J. AM. CHEM. SOC. VOL. 132, NO. 25, 2010 8711

A R T I C L E S

Schwefel et al.

Figure 5. Comparison of binding modes of divalent ligand 23 (cyan, site pair B2C1-C1B2) and tetravalent cyclic peptide ligand 2 (white, site pair

B1C2-C2B1). The ligands are shown in ball-and-stick representation, with carbon atoms colored cyan and white, respectively, nitrogens blue, and oxygens

red. We compare ligands at different binding site pairs because 23 bound to site pair B1C2-C2B1 is compromised by crystallographic contacts (cf. Figure

3A). Nevertheless, this comparison is valid, because site pairs B1C2-C2B1 and B2C1-C1B2 are structurally equivalent due to pseudo-twofold rotational

symmetry of the WGA dimer.

modeling of the cyclic peptides is impossible indicating that a

multiplicity of conformations ﬂattens the electron density map.

The binding mode of the ligand’s GlcNAc-carbamate moieties

is identical to the binding mode of glycosyl carbamate 20 and

the GlcNAc-carbamate moieties of divalent compound 23. There

are no interactions between the protein and the carbamate groups

or the peptide part of the ligand.

28 (Scheme 5) and subjected it to the ELLA. An IC₅₀value of

191 µM (rel. potency 120 over GlcNAc, 42 over 20) was

determined, being even worse than that of divalent ligand 23

(Table 1). Dipeptide 28 differs from tetravalent ligand 2 not

only by lacking two sugar residues, but also in being more

ﬂexible. Possibly, the conformation of the cyclic peptide scaffold

orients the side chains of the glycosylated D-Dab residues 4

and 5 in a way facilitating bivalent binding to WGA. To test

this hypothesis, we determined the conformation of 2 in solution

by NMR analysis and compared this structure with the fragment

of 2 that is visible in the crystal structure.

Comparison of Linker Lengths in Spanning Ligands. The

divalent ligands 21, 22, and 23 differ in their spacer length with

10, 17, and 18 single bonds, respectively, between the anomeric

oxygens. The somewhat meandering appearance of the linker

region seen at the high afﬁnity binding site pairs in complex

WGA3∼23 and the missing electron density for the linker at

the low afﬁnity site pairs suggest that the linker is slightly too

long. This is supported by the signiﬁcantly higher binding

potency of compound 22, whose linker is one bond length

shorter than in compound 23 (Table 1). Superposition of the

structures of the linker of bound divalent 23 and the bridging

dipeptide subunit of compound 2 reveals that the dipeptide ﬁts

even better to the architecture of the binding site pair (Figure

5). This could contribute to the increased afﬁnity of the cyclic

peptide ligand compared to the divalent ligands 23 and 22. Two

further glycosylated D-Dab residues are available in positions

2 and 7 of tetravalent peptide 2, which are separated from the

WGA-binding sugar pair (residues 4 and 5) by D-valine residues.

In the preformed environment of the cross-linked crystal, these

cannot access binding sites. In solution, however, they may

further enhance the binding afﬁnity of 2 by interacting with

additional WGA molecules. Indeed, addition of the tetravalent

cyclic peptide ligand 2 to WGA solutions resulted in precipita-

tion of the protein which indicates cross-linking of WGA dimers

to higher oligomers.

According to Table 1, the shortest of the divalent ligands,

21, apparently does not allow high-afﬁnity chelating binding

to WGA. This is supported by molecular modeling studies.

Although it was possible to ﬁt 21 in the crystallographically

determined WGA structure in a mode bridging adjacent high-

afﬁnity sites without obvious conformational strain, even short

molecular dynamic simulations (Tripos 5.2 forceﬁeld⁴⁴) resulted

in dissociation of one of the GlcNAc residues. Under the same

conditions, the positions of the GlcNAc moieties of divalently

bound 23 changed only marginally over a period of 50 ps (data

not shown).

Conformational Analysis of 2 by NMR Spectroscopy. The

conformational analysis of 2 was based on a complete NMR

resonance assignment which was obtained from homonuclear

DQF-COSY, TOCSY, and NOESY, as well as heteronuclear

¹H-¹³C-HSQC spectra (Table S1). TOCSY amide traces of 2

(Figure 6A) are well dispersed leading to unambiguous assign-

ments of all backbone H^Nand H^Rprotons as well as side chain

resonances of Lys-1, Val-3, Val-6, Glu-8, and ꢀ-Ala-9. How-

ever, the side chain resonances of the D-Dab amino acids and

the signals of the four GlcNAc residues are heavily overlapped

such that in most cases only average values could be determined.

This indicates that the sugar-bearing D-Dab side chains and the

sugar moieties themselves are, in the absence of WGA, solvent

accessible, largely unrestricted, and sample many conformations

which leads to similar chemical shifts for all four glycosylated

side chains. A similar absence of chemical shift dispersion is

also observed for the side chains of Lys-1 and Glu-8 that form

the lactam bridge of the cyclopeptide suggesting that this part

of the molecule has to be considered ﬂexible, as well. In contrast

to this, the chemical shift dispersion of the backbone resonances

of 2 allowed for extracting a number of structural restraints.

A total of 25 structurally relevant, unambiguous interproton

distances were derived from analyzing cross peak volumes of

2D NOESY spectra recorded with a mixing time of 50 ms (Table

S2). Four of these were long-range (i.e., nonsequential) NOE

contacts. In addition, four ambiguous distance constraints could

be identiﬁed. Furthermore, the 3D structure of 2 is deﬁned by

four φ and ﬁve ꢁ₁torsion angle constraints that were obtained

from ³J_HNHRand ³J_HRHꢀcoupling constants (Tables S3 and S4).

In the case of φ torsion angles, direct J-coupling restraints were

3

employed only if the J_HNHRcoupling was >8 Hz resulting in

only two similar solutions of the Karplus equation.⁴⁵According

to the presence of, in each case, one large and one small ³J_HRHꢀ

coupling constant with sums of 15-18 Hz, the ꢁ₁angles of the

D-Dab and Glu amino acids were restrained such that the g^-

rotamer was excluded in the D-Dab residues and the g⁺rotamer

To test whether the substructure of 2 that is visible in the

crystal structure (i.e., residues 4 and 5) alone is sufﬁcient for

high-afﬁnity binding to WGA, we synthesized divalent dipeptide

(44) Clark, M.; Cramer, R. D., III; Van Opdenbosch, N. J. Comput. Chem.

1989, 10, 982–1012.

(45) Karplus, M. J. Am. Chem. Soc. 1963, 85, 2870–2871.

9

8712 J. AM. CHEM. SOC. VOL. 132, NO. 25, 2010

Multivalent Binding to Wheat Germ Agglutinin

A R T I C L E S

Figure 6. (A) Amide region of a 600 MHz TOCSY spectrum of 2 in H₂O/D₂O at 300 K. (B) Ensemble of the dominant family of conformers of 2. (C)

Orientations and accessible volume of sugar-substituted side chains of the ensemble (blue, D-Dab-2; red, D-Dab-4; green, D-Dab-5; magenta, D-Dab-7).

Table 3. Structure Determination Statistics of Neoglycopeptide 2

A. NMR restraints

Total unambiguous distance restraints^a

25

21

4

Sequential (i,i+1)

Long-range (|i-j| > 1)

Total ambiguous distance restraints

4

Total dihedral angle restraints

9

Φ

ꢁ_I

4

5

B. Ensemble statistics (20 structures)

Violation analysis

Maximum distance violation (Å)

Maximum dihedral angle violation (deg)

Energies

0.28

<5

Mean restraint violation energy (kcal mol^-1

)

2.534

-510.18

Mean AMBER energy (kcal mol^-1

)

Mean deviation from ideal covalent geometry

Bond lengths (Å)

Bond angles (deg.)

0.0121

2.18

conformational family

I

II

III

Number of conformers

9

6

5

Average rmsd from the conformer with lowest restraint violation energy

within each conformational family (backbone C, N, and O atoms) (Å)

0.65

1.12

0.71

^aDistance restraints that refer to nonstereospeciﬁcally assigned protons are included here.

in the Glu residue. Long range and dihedral angle constraints

are summarized in Figure S5. We used the dedicated biopolymer

force ﬁeld of the AMBER software package⁴⁶that provides

highly versatile options for NMR-restrained structure calcula-

tions. Twenty conformers with low combined restraint violation

were selected for further analysis. The statistics of the ﬁnal

structural ensemble is summarized in Table 3.

of D-amino acids are mirror images of those of L-amino acids

(Figure S6). Neoglycopeptide 2 adopts a triangular shape with

turns at D-Dab-2, D-Dab-5, and Glu-8. The region of highest

structural similarity among all three families of conformers

appears to range from D-Dab-2 to D-Val-6. Here, dihedral an-

gle combinations are present that correspond to R_R, ꢀ, ꢀ, R_R,

and ꢀ secondary structure, respectively, of peptides com-

posed of L-amino acids. Stretches with the same secondary

structure extending over more than two residues as well as turn

conformations with distinct hydrogen bonding pattern are

lacking.

Comparison to X-ray Structure. The X-ray structure of

WGA1 in complex with neoglycopeptide 2 contains little

information about the conformation of the cyclopeptide back-

bone. Only the fragment ranging from the amide nitrogen of

D-Dab-4 to the carbonyl of D-Dab-5 is resolved in the electron

density map. On the other hand, the solution structure does not

report on well-deﬁned conformations of the sugar-decorated side

chains. They are conformationally unrestricted when the protein

is not present (Figure 6C). Thus, the two methods provide

complementary information about the structure. To ﬁnd out

Description of the Structural Ensemble. The ensemble of

conformers of neoglycopeptide 2 can be grouped into three

conformational families consisting of 9, 6, and 5 conformers,

respectively, according to backbone rmsd values with respect

to conformers with lowest restraint violation energy. The

conformers of the dominant conformational family superimpose

with an average rmsd of 0.65 Å with the member having the

lowest restraint violations (Figure 6B). All amino acids adopt

allowed backbone dihedral angles when considering that the

energetically favorable regions in the Ramachandran diagram

(46) Case, D. A.; Cheatham, T. E., 3rd; Darden, T.; Gohlke, H.; Luo, R.;

Merz, K. M., Jr.; Onufriev, A.; Simmerling, C.; Wang, B.; Woods,

R. J. J. Comput. Chem. 2005, 26, 1668–1688.

9

J. AM. CHEM. SOC. VOL. 132, NO. 25, 2010 8713

A R T I C L E S

Schwefel et al.

sites). Employing a GlcNAc carbamate motif with R-conﬁgu-

ration and by systematic variation of the spacer length, we were

able to identify divalent ligands with unprecedented high WGA

binding potency. The best divalent ligand (22) has an IC₅₀value

of 9.8 µM corresponding to a relative potency of 2350 (1170

on a valency-corrected basis, i.e., per mol sugar contained)

compared to free GlcNAc. This is the highest inhibitory activity

of any divalent WGA ligand yet reported.^28,47X-ray crystal-

lography of the complex of WGA and the second best, closely

related ligand 23 explains this activity. Four molecules of 23

simultaneously bind to WGA with each ligand bridging adjacent

binding sites which are approximately 14 Å apart from each

other (distance between the anomeric centers of bound GlcNAc

moieties). To the best of our knowledge, this is the ﬁrst crystal

structure of a divalent lectin ligand binding in a chelating

fashion. From this structure, it also becomes clear that the

R-glycosidic linkage is a key feature of the spanning ligands.

The axial glycosidic bonds of bound GlcNAc pairs point to each

other facilitating the construction of a suitable linker. With

ꢀ-glycosidic linkages, a considerably longer linker would be

required. Although it has been shown that WGA can bind to

simple R- and ꢀ-glycosides of GlcNAc with similar afﬁnity (cf.

Table 1 and literature³¹), our X-ray structure suggests such

ꢀ-glycosidically linked divalent ligands to be less favorable to

high-afﬁnity binding for entropic reasons.

Figure 7. Superposition of the dominant conformational family of

neoglycopeptide 2 determined by solution NMR spectroscopy (atom color)

onto the resolved atoms of the ligand in the crystal (orange). The heavy

atoms of the sugar-bearing side chains and the D-Dab-4-D-Dab-5 peptide

backbone were ﬁtted by minimizing their distances to the X-ray structure

in an MD simulation (see text).

Binding to the four low-afﬁnity (secondary) WGA binding

sites has been reported to be too weak to be detectable in

solution. Low occupancy of the secondary binding sites was

also observed in the complex of WGA1 and monovalent propyl

carbamate 20. The high binding afﬁnity of divalent ligand 23,

on the other hand, explains why in the complex WGA3∼23 all

binding sites including the secondary ones are occupied. Thus,

multivalent lectin ligands with suitable spacers can be a tool

for generating crystal structures of multivalent lectins with low-

afﬁnity binding sites occupied.

With an IC₅₀value of 0.9 µM, the binding potency of the

tetravalent neoglycopeptide 2 is 25 500 times higher than that

of GlcNAc (6400 times per contained sugar). Comparison of

the X-ray structure and the solution NMR structure as well as

comparison of the IC₅₀values of dipeptide 28 and divalent ligand

23 (Table 1) suggests that the backbone conformation of the

fragment D-Dab-4-D-Dab-5 is already preorganized in solution

to adopt a conformation suitable for bridging the high-afﬁnity

binding site pairs. However, the two additional GlcNAc residues

contained in 2 allow for additional binding modes (spanning of

binding sites by alternative GlcNAc pairs or cross-linking of

several WGA molecules) that are likely to sum up to the

observed high afﬁnity.

whether the solution conformation of 2, as determined by NMR,

is compatible with its WGA-bound conformation determined

by X-ray crystallography, we superimposed both independent

structures.

The solution structure was annealed onto the crystal structure

by employing distance restraints (1 kcal Å^-2) from all resolved

C, N, and O atoms of the ligand fragment to the corresponding

atoms of the NMR structure. These distances were minimized

during short molecular dynamics simulations and subsequent

energy minimization. The backbone conformation of the cy-

clopeptide was preserved during this ﬁtting procedure by strong

dihedral angle restraints (20 kcal deg^-2). The members of the

dominant NMR ensemble achieve an average rmsd of 1.31 Å

without signiﬁcant conformational strain of the GlcNAc deco-

rated side chains documenting that the decisive fragment can

adopt a similar conformation in solution and in complex with

WGA (Figure 7). This implies that cyclopeptide 2 is already

preorganized in solution in a conformation suitable for the

interaction of two GlcNAc residues with the protein thereby

contributing to the high afﬁnity of 2.

The structures of the second family of conformers superim-

pose with the resolved fragment in the crystal with an average

rmsd of 1.22 Å again showing natural side chain conformations.

The third conformational family would also allow a similar

binding mode as found in the crystal (rmsd 1.30 Å); however,

one member of this ensemble would be affected by steric

clashing into the protein.

The high-resolution crystal structure of WGA in complex with

four molecules of a divalent ligand reveals that all binding sites

of the WGA dimer are simultaneously functional. This is now

the basis for the design of future high-afﬁnity di- and higher

valent WGA ligands. The structure immediately suggests that

tetravalent ligands composed of two divalent units of type 22

interconnected by a linker can be envisioned that are either able

to bind to four binding sites of a single WGA dimer resulting

Conclusions

In conclusion, we have synthesized a series of multivalent

WGA ligands and studied their interaction with the protein by

an enzyme-linked lectin assay and X-ray crystallography. The

solution conformation of one ligand was determined by NMR

spectroscopy. WGA is a plant lectin that differs from most

lectins by the fact that it has several binding sites per subunit

that are distributed over the whole protein. The WGA dimer

accommodates eight independent binding sites (four unique

(47) (a) Zanini, D.; Roy, R. Bioconjugate Chem. 1997, 8, 187–192. (b)

Furuike, T.; Aiba, S.; Nishimura, S.-I. Tetrahedron 2000, 56, 9909–

9915. (c) Consoli, G. M. L.; Cunsolo, F.; Geraci, C.; Sgarlata, V. Org.

Lett. 2004, 6, 4163–4166. (d) Yamada, A.; Hatano, K.; Koyama, T.;

Matsuoka, K.; Takahashi, N.; Hidari, K. I. P. J.; Suzuki, T.; Suzuki,

Y.; Terunuma, D. Bioorg. Med. Chem. 2007, 15, 1606–1614. (e)

Misawa, Y.; Akimoto, T.; Amarume, S.; Murata, T.; Usui, T.

J. Biochem. 2008, 143, 21–30.

9

8714 J. AM. CHEM. SOC. VOL. 132, NO. 25, 2010

Multivalent Binding to Wheat Germ Agglutinin

A R T I C L E S

NovaSyn TG Sieber resin (Novabiochem, Darmstadt, Germany)

(abbreviation: Sieber-TG) using an ABI 433A peptide synthesizer

(Applied Biosystems, Foster City, CA) applying the FastMoc-

Chemistry data ﬁle (“PrPK Fmoc Depro/Single”). The loading of

the resin after immobilization of the ﬁrst amino acid was 0.2 mmol

g^-1. Peptide couplings were carried out in 1-methyl-2-pyrrolidone

(NMP) with 5 equiv of Fmoc-amino acid and activation by HBTU

(4.75 equiv), HOBt (4.75 equiv), and EtNi-Pr₂(10 equiv). In the

last coupling step, Boc-Lys(Aloc)-OH was coupled.

cyclo[Boc-Lys-D-Dab(Ddv)-D-Val-D-Dab(Ddv)-D-Dab(Ddv)-

D-Val-D-Dab(Ddv)-Glu]-ꢀ-Ala- Sieber-TG (8). After having been

swollen in CH₂Cl₂for 10 min, resin 7 was shaken with a solution

of [Pd(PPh₃)₄] (1.6 equiv) and BH₃·HNMe₂(30 equiv) in CH₂Cl₂

(2 × 20 min) and subsequently washed with DMF (5 × 2 min),

0.5% sodium diethyldithiocarbamate solution in DMF (3 × 1 min),

DMF (10 × 1 min), and CH₂Cl₂(4 × 2 min). Then, the resin was

treated with 5% HOBt in DMF (3 × 2 min) and washed with DMF

(5 × 1 min) and NMP (2 × 1 min). Side-chain cyclization was

induced by addition of a solution of HOBt (6 equiv), HBTU (4

equiv), and EtNi-Pr₂(8 equiv) in NMP. This solution was replaced

several times until Kaiser test and TNBS test indicated completion

of the reaction followed by washing the resin with DMF (6 × 2

min) and CH₂Cl₂(3 × 2 min).

in even better binding afﬁnity or could, for example, ﬁnd

application as lectin cross-linker. It is also obvious that even

subtle changes of the linker structure (cf. 22/23) can have a

signiﬁcant impact on the binding afﬁnity. Thus, combinatorial

approaches such as our spatial screening procedure^20bthat probe

for linker length and ﬂexibility are valuable tools for the

identiﬁcation of optimized multivalent ligands of proteins with

unknown 3D structure.

Connecting ligands that can bind to different protein binding

sites by a suitable linker is an established strategy in fragment-

based drug discovery aiming at improved association con-

stants.⁴⁸Our ﬁndings provide clear experimental evidence for

the validity of this approach and show that high-afﬁnity binding

to a protein can be achieved by carefully designed ligands that

are able to bridge several pairs of binding sites.

Experimental Section

General Methods. Solvents were puriﬁed by distillation and

dried by usual procedures. Thin-layer chromatography (TLC) was

performed on Merck silica gel 60 F₂₅₄aluminum sheets. Compound

spots were visualized by UV light and immersion in the appropriate

stain followed by heating. Preparative ﬂash column chromatography

(FC) was carried out on Merck silica gel 60 (40-63 µm). ¹H NMR

and ¹³C NMR spectra were recorded at 300 K on Bruker AC 250,

Bruker Avance 400, and Bruker Avance DRX 600 instruments.

¹H chemical shifts are reported in ppm relative to the signal from

residual protic solvent (CDCl₃, 7.26 ppm; D₂O, 4.67 ppm). ¹³C

chemical shifts are referenced to the solvent signal (CDCl₃: 77.0

ppm) or in case of D₂O to internal standard CD₃CN (1.3 ppm).

Assignments of proton and carbon signals were carried out with

the aid of DQF-COSY, HMQC, and ROESY experiments. High

resolution ESI mass spectra were obtained on a Bruker micrOTOF

II instrument (positive mode). Elemental analyses were per-

formed by the microanalytical facility at the Universita¨t Konstanz.

p-Nitrophenoxycarbonyl 2-Acetamido-3,4,6-tri-O-acetyl-2-

deoxy-r-D-glucopyranoside (5). Compound 4^30,49(1.04 g, 3 mmol)

was dissolved in dry CH₂Cl₂(10 mL) and cooled to 0 °C. After

addition of Et₃N (832 µL, 6 mmol), a solution of p-nitrophenyl

chloroformate (1.33 g, 6.6 mmol) in dry CH₂Cl₂(1 mL) was added

dropwise upon which a precipitate formed. After a reaction time

of 3 h, the solvent was evaporated and the remainder dissolved in

ethyl acetate and washed with 10% aq citric acid and water. The

combined aqueous phases were re-extracted with ethyl acetate. The

combined organic phases were dried (MgSO₄) and concentrated.

Short column FC (ethyl acetate/petroleum ether 2:1, then ethyl

acetate) gave 5 (1.41 g, 92%) as white solid. R_f) 0.51 (ethyl

Cyclic Neoglycopeptide (2). Resin 8 was treated with hydrazine

hydrate/DMF (4:96) (6 × 5 min) and washed with DMF (10 × 1

min). A solution of active carbonate 5 (12 equiv) and EtNi-Pr₂(12

equiv) in NMP was added, and after completion of the coupling

step (reaction control by Kaiser test and TNBS test), the resin was

washed with DMF (10 × 1 min) and CH₂Cl₂(5 × 2 min). To cleave

acetylated cyclic glycopeptide 9 from the solid phase, the resin was

treated with TFA/i-Pr₃SiH/CH₂Cl₂(1:1:98) (1 × 10 min, 3 × 2

min) and with TFA/i-Pr₃SiH/Mix (1:1:98) (Mix: CH₂Cl₂/F₃CCH₂OH

3:1) (1 × 5 min, 2 × 2 min) and thoroughly washed with

F₃CCH₂OH, CH₂Cl₂/F₃CCH₂OH (3:1), CH₂Cl₂, MeOH, and ﬁnally

CH₂Cl₂(each 2 × 2 min). The combined ﬁltrates were neutralized

with pyridine and concentrated under vacuum. Analytical RP-HPLC

(Knauer Nucleosil C18, 4 × 250 mm, ﬂow: 1 mL min^-1, 5-100%

MeCN in H₂O/0.1% TFA over 30 min) of 9: t_R) 16.2 min.

Acetylated glycopeptide 9 was deprotected by treatment with

NaOMe (27 mM) in dry MeOH/CHCl₃(1:1) for 3.5 h. After

neutralization with weakly acidic ion-exchange resin (Amberlite

IRC-86) and ﬁltration, the solvent was removed under vacuum. The

remaining ion-exchange resin was washed with MeOH and water

and the combined ﬁltrates were lyophilized and puriﬁed by RP-

HPLC (Knauer Eurospher C18, 16 × 250 mm, ﬂow ) 9.6 mL

min^-1, 10-35% MeCN in H₂O/0.1% TFA over 30 min; t_R) 16.6

min) to give cyclic neoglycopeptide 2 in a yield of 15% based on

initial resin loading. NMR: see Table S1; MALDI-MS (matrix:

R-cyano-4-hydroxycinnamic acid): calcd for C₈₁H₁₃₅N₁₉O₄₀m/z

2036.9 [M + Na]⁺, 2052.9 [M + K]⁺; found, 2036.8, 2052.8.

1

acetate); H NMR (CDCl₃, 400.1 MHz): δ 8.34-8.32 (m, 2H,

arenes), 7.48-7.46 (m, 2H, arenes), 6.20 (d, J ) 3.6 Hz, 1H, H-1),

5.82 (d, J ) 8.6 Hz, 1H, NH), 5.36-5.26 (m, 2H, H-3, H-4), 4.56

(ddd, J ) 3.6, 8.5, 10.6 Hz, 1H, H-2), 4.33 (dd, J ) 4.0, 12.5 Hz,

1H, H-6a), 4.19-4.14 (m, 2H, H-6b, H-5), 2.12 (s, 3H, C(O)CH₃),

2.10 (s, 3H, C(O)CH₃), 2.08 (s, 3H, C(O)CH₃), 2.01 (s, 3H,

C(O)CH₃); ¹³C NMR (CDCl₃, 150.9 MHz): δ 171.7, 170.6, 170.2,

169.0 (4 × C(O)CH₃), 154.9 (O-C(O)-O), 150.5 (C^arene), 145.6

(C^arene), 125.4 (HC^arene), 121.4 (HC^arene), 95.7 (C-1), 70.2, 70.1, 67.1

(C-3, C-4, C-5), 61.2 (C-6), 51.3 (C-2), 23.0 (C(O)CH₃), 20.7 (2

× C(O)CH₃), 20.5 (C(O)CH₃); elemental analysis calcd (%) for

C₂₁H₂₄N₂O₁₃(512.42): C, 49.22; H, 4.72; N, 5.47. Found: C, 49.49;

H, 4.91; N, 5.22.

General Procedure 1 (GP 1): Coupling of Activated

Carbonate 5 to Amino-Substituted Scaffolds. p-Nitrophenyl

carbonate 5 (1.1 equiv per amino group) was dissolved in dry

CH₂Cl₂(8-15 mL per mmol 5). Addition of the amine and Hu¨nig’s

Base (1 equiv base per amino group) afforded a yellow colored

solution. The reaction mixture was stirred at room temperature until

the reaction was complete according to TLC. After evaporation of

the solvent, the residue was puriﬁed by FC.

General Procedure 2 (GP 2): De-O-acetylation Using

MeOH/EtNMe₂. The acetylated compound was dissolved in

MeOH/EtNMe₂5:1 and stirred at room temperature until deacety-

lation was complete according to TLC. Evaporation of the solvent

gave the product in quantitative yield.

Boc-Lys(Aloc)-D-Dab(Ddv)-D-Val-D-Dab(Ddv)-D-Dab(Ddv)-

D-Val-D-Dab(Ddv)-Glu(OAll)-ꢀ-Ala-NH-Sieber-TG (7). Solid-

phase peptide synthesis was carried out in 20 µmol scale on

(48) (a) Fragment-based Approaches in Drug DiscoVery; Jahnke, W.;

Erlanson, D. A., Eds.; Wiley-VCH: Weinheim, 2006; (b) Kramer,

R. H.; Karpen, J. W. Nature 1998, 395, 710–713. (c) Shuker, S. B.;

Hajduk, P. J.; Meadows, R. P.; Fesik, S. W. Science 1996, 274, 1531–

1534.

General Procedure 3 (GP 3): De-O-acetylation Using

NaOMe/MeOH. The acetylated compound was dissolved in MeOH

and catalytic amounts of a solution of NaOMe in MeOH were

added. After stirring at room temperature until the deacetylation

was complete, the reaction mixture was neutralized using acidic

(49) Jha, R.; Davis, J. T. Carbohydr. Res. 1995, 277, 125–134.

9

J. AM. CHEM. SOC. VOL. 132, NO. 25, 2010 8715

A R T I C L E S

Schwefel et al.

ion-exchange resin (Amberlite IRC-86, H⁺form). Filtration and

evaporation of the solvent gave the product.

NH), 5.22-5.19 (m, 4H; H-3, H-4), 4.53-4.51 (m, 2H; H-2), 4.24

(dd, J ) 12.4, 3.7 Hz, 2H; H-6a), 4.09-4.02 (m, 4H; H-6b, H-5),

3.68-3.57 (m, 12H; CH₂CH₂, NCH₂CH₂CH₂O), 3.33-3.31 (q, J

) 6.2 Hz, 4H; NCH₂CH₂CH₂O), 2.08 (s, 6H; C(O)CH₃), 2.03 (s,

6H; C(O)CH₃), 2.02 (s, 6H; C(O)CH₃), 1.93 (s, 6H; C(O)CH₃),

1.83-1.80 (m, 4H; NCH₂CH₂CH₂O); ¹³C NMR (CDCl₃, 150.9

MHz): δ 171.5, 170.8, 170.1, 169.1 (C(O)CH₃), 153.9 (OC(O)NH),

91.6 (C-1), 70.9 (C-3), 70.3 (CH₂CH₂), 70.0 (CH₂CH₂), 69.4 (C-

5), 69.3 (NCH₂CH₂CH₂O), 67.7 (C-4), 61.6 (C-6), 50.8 (C-2), 39.3

(NCH₂CH₂CH₂O), 29.2 (NCH₂CH₂CH₂O), 22.9, 20.7, 20.7, 20.6

(C(O)CH₃); elemental analysis calcd (%) for C₄₀H₆₂N₄O₂₃(966.93):

C, 49.69; H, 6.46; N, 5.79. Found: C, 49.41; H, 6.47; N, 5.48.

Tris-(2-(2-acetamido-3,4,6-tri-O-acetyl-2-deoxy-r-D-glucopy-

ranosyloxycarbonylamino)-ethyl)-amine (19). According to GP

1, 5 (338 mg, 0.66 mmol), CH₂Cl₂(5 mL), tris-(2-aminoethyl)-

amine (30 µL, 0.2 mmol), and EtN(i-Pr)₂(105 µL, 0.6 mmol) were

reacted for 2.5 h. FC (CH₂Cl₂/MeOH 9:1) gave 19 (220 mg, 88%)

N-(2-Acetamido-3,4,6-tri-O-acetyl-2-deoxy-r-D-glucopyran-

osyloxycarbonyl)-propylamine (15). According to GP 1, 5 (50

mg, 0.098 mmol), CH₂Cl₂(1.5 mL), propylamine (7.3 µL, 0.089

mmol), and EtN(i-Pr)₂(15 µL, 0.089 mmol) were reacted for 1.5 h.

FC (EtOAc) gave 15 (34 mg, 89%) as a white solid. R_f) 0.16

1

(EtOAc); H NMR (CDCl₃, 400.1 MHz): δ 6.06 (d, J ) 3.4 Hz,

1H, H-1), 5.70 (br s, 1H, NHAc), 5.23-5.15 (m, 2H, H-3, H-4),

5.09 (br s, 1H, NH), 4.50-4.45 (m, 1H, H-2), 4.23 (dd, J ) 12.4,

3.9 Hz, 1H, H-6a), 4.06 (d, J ) 12.6 Hz, 1H, H-6b), 4.02.3.99 (m,

1H, H-5), 3.24-3.12 (m, 2H, CH₂CH₂CH₃), 2.06 (s, 3H, C(O)CH₃),

2.02 (s, 3H, C(O)CH₃), 2.01 (s, 3H, C(O)CH₃), 1.93 (s, 3H,

C(O)CH₃), 1.56 (sext, J ) 7.3 Hz, 2H, CH₂CH₂CH₃), 0.94 (t, J )

7.4 Hz, 3H, CH₂CH₂CH₃);¹³C NMR (CDCl₃, 100.6 MHz): δ 171.5,

170.7, 170.0, 169.1 (4 × C(O)CH₃), 153.7 (OC(O)NH), 91.4 (C-

1), 70.7 (C-3 or C-4), 69.4 (C-5), 67.6 (C-4 or C-3), 61.6 (C-6),

50.9 (C-2), 42.9 (CH₂CH₂CH₃), 23.0 (CH₂CH₂CH₃), 22.9, 20.7,

20.6, 20.5 (C(O)CH₃), 11.1 (CH₂CH₂CH₃); elemental analysis calcd

(%) for C₁₈H₂₈N₂O₁₀(432.42): C, 50.00; H, 6.53; N, 6.48. Found:

C, 50.08; H, 6.49; N, 6.49.

1

as a white solid. R_f) 0.26 (CH₂Cl₂/MeOH 9:1); H NMR (600.1

MHz, CDCl₃): δ 6.75 (d, J ) 8.7 Hz, 3H; NHAc), 6.25 (d, J ) 3.1

Hz, 3H; H-1), 5.84 (br s, 3H; NH), 5.26 (t, J ) 10.2 Hz, 3H; H-3),

5.19 (t, J ) 9.8 Hz, 3H; H-4), 4.58-4.54 (m, 3H; H-2), 4.24 (dd,

J ) 12.3, 3.5 Hz, 3H; H-6a), 4.12-4.10 (m, 6H; H-6b, H-5), 3.28

(br s, 6H; CH₂CH₂), 2.67-2.58 (m, 6H; CH₂CH₂), 2.07 (s, 9H;

C(O)CH₃), 2.04 (s, 9H; C(O)CH₃), 2.03 (s, 9H; C(O)CH₃), 1.94

(s, 9H; C(O)CH₃);¹³C NMR (CDCl₃, 150.9 MHz): δ 171.0, 170.7,

170.5, 169.5 (C(O)CH₃), 155.1 (OC(O)NH), 91.9 (C-1), 70.4 (C-

3), 69.3 (C-5), 68.2 (C-4), 61.8 (C-6), 55.8 (CH₂CH₂), 50.8 (C-2),

40.1 (CH₂CH₂), 22.8, 20.7, 20.6, 20.6 (C(O)CH₃); elemental

analysis calcd (%) for C₅₁H₇₅N₇O₃₀(1266.17): C, 48.38; H, 5.97;

N, 7.74. Found: C, 48.28; H, 6.01; N, 7.30.

1,5-Bis-(2-acetamido-3,4,6-tri-O-acetyl-2-deoxy-r-D-glucopy-

ranosyloxycarbonylamino)-3-oxa-pentane (16). According to GP

1, 5 (307 mg, 0.6 mmol), CH₂Cl₂(5.5 mL), 1,5-diamino-3-

oxapentane (29 µL, 0.27 mmol), and EtN(i-Pr)₂(94 µL, 0.55 mmol)

were reacted for 2 h. FC (CH₂Cl₂/MeOH 20:1) gave 16 (230 mg,

100%) as a white solid. R_f) 0.29 (CH₂Cl₂/MeOH 15:1); ¹H NMR

(CDCl₃, 400.1 MHz): δ 6.85 (d, J ) 6.8 Hz, 2H, NHAc), 6.15 (d,

J ) 3.5 Hz, 2H, H-1), 6.10 (t, J ) 5.3 Hz, 2H, NH), 5.20 (t, J )

10.0 Hz, 2H, H-3), 5.13 (t, J ) 9.6 Hz, 2H, H-4), 4.54 (dt, J )

9.7, 3.7 Hz, 2H, H-2), 4.21 (dd, J ) 12.4, 3.5 Hz, 2H, H-6a),

4.10-4.07 (m, 2H, H-5), 4.03 (d, J ) 12.2 Hz, 2H, H-6b), 3.58 (t,

J ) 4.4 Hz, 4H, NHCH₂CH₂O), 3.44-3.27 (m, 4H, NHCH₂CH₂O),

2.01 (s, 6H, C(O)CH₃), 1.97 (s, 12H, C(O)CH₃), 1.91 (s, 6H,

C(O)CH₃);¹³C NMR (CDCl₃, 100.6 MHz): δ 170.8, 170.6, 170.5,

169.4 (C(O)CH₃), 154.9 (OC(O)NH), 91.4 (C-1), 70.4 (C-3), 69.6

(NHCH₂CH₂O), 69.2 (C-5), 68.0 (C-4), 61.8 (C-6), 50.7 (C-2), 40.4

(NHCH₂CH₂O), 22.6, 20.6, 20.5, 20.4 (C(O)CH₃); elemental

analysis calcd (%) for C₃₄H₅₀N₄O₂₁(850.78): C, 48.00; H, 5.92;

N, 6.59. Found: C, 47.62; H, 6.07; N, 6.44.

1,4-Bis-(3-(2-acetamido-3,4,6-tri-O-acetyl-2-deoxy-r-D-glu-

copyranosyloxycarbonylamino)-propoxy)-butane (17). According

to GP 1, 5 (225 mg, 0.44 mmol), CH₂Cl₂(5 mL), 1,4-bis-(3-

aminopropoxy)-butane (43 µL, 0.2 mmol), and EtN(i-Pr)₂(70 µL,

0.4 mmol) were reacted for 3 h. FC (CH₂Cl₂/MeOH 15:1) gave 17

(185 mg, 97%) as a white solid. R_f) 0.20 (CH₂Cl₂/MeOH 15:1);

¹H NMR (CDCl₃, 600.1 MHz): δ 6.08 (d, J ) 3.5 Hz, 2H; H-1),

5.68 (d, J ) 9.3 Hz, 2H; NHAc), 5.44 (t, J ) 5.3 Hz, 2H; NH),

5.23-5.18 (m, 4H; H-4, H-3), 4.53-4.50 (m, 2H; H-2), 4.25 (dd,

J ) 12.4, 3.6 Hz, 2H; H-6a), 4.08 (dd, J ) 12.3, 1.6 Hz, 2H; H-6b),

4.01-4.00 (m, 2H; H-5), 3.53-3.50 (m, 4H; CH₂CH₂CH₂), 3.47

(br s, 4H; CH₂CH₂), 3.35-3.33 (m, 4H; CH₂CH₂CH₂), 2.09 (s, 6H;

C(O)CH₃), 2.06 (s, 6H; C(O)CH₃), 2.03 (s, 6H; C(O)CH₃), 1.94

(s, 6H; C(O)CH₃), 1.84-1.81 (m, 4H; CH₂CH₂CH₂), 1.67 (br s,

4H; CH₂CH₂); ¹³C NMR (CDCl₃, 150.9 MHz): δ 171.6, 170.8,

169.9, 169.1 (C(O)CH₃), 153.7 (OC(O)NH), 91.5 (C-1), 70.8 (C-3

or C-4), 70.7 (CH₂CH₂), 69.4 (C-5), 69.0 (CH₂CH₂CH₂), 67.6 (C-4

or C-3), 61.5 (C-6), 50.9 (C-2), 39.5 (CH₂CH₂CH₂), 29.4

(CH₂CH₂CH₂), 26.4 (CH₂CH₂), 23.1, 20.7, 20.7, 20.6 (C(O)CH₃);

elemental analysis calcd (%) for C₄₀H₆₂N₄O₂₂(950.93): C, 50.52;

H, 6.57; N, 5.89. Found: C, 50.20; H, 6.60; N, 5.85.

(2-Acetamido-2-deoxy-r-D-glucopyranosyloxycarbonyl)-pro-

pylamine (20). According to GP 2, 15 (180 mg, 0.375 mmol) was

dissolved in MeOH/EtMe₂N 5:1 and reacted for 3 days. Compound

20 was obtained in quantitative yield as a white solid. R_f) 0.58

1

(MeCN/H₂O 4:1); H NMR (600.1 MHz, D₂O): δ 5.87 (br s, 1H;

H-1), 3.97-3.96 (m, 1H; H-2), 3.76-3.67 (m, 4H; H-3, H-5, H-6a,

H-6b), 3.50 (t, J ) 9.2 Hz, 1H; H-4), 3.04 (br s, 2H; CH₂CH₂CH₃),

1.95 (s, 3H; C(O)CH₃), 1.44 (q, J ) 6.6 Hz, 2H; CH₂CH₂CH₃),

0.82 (t, J ) 6.8 Hz, 3H; CH₂CH₂CH₃); ¹³C NMR (150.9 MHz,

D₂O): δ 175.4 (C(O)CH₃), 157.4 (OC(O)NH), 92.3 (C-1), 74.6 (C-3

or C-5), 71.5 (C-5 or C-3), 70.3 (C-4), 61.1 (C-6), 53.5 (C-2), 43.1

(CH₂CH₂CH₃), 23.0 (CH₂CH₂CH₃), 22.6 (C(O)CH₃), 11.3

(CH₂CH₂CH₃); HR-MS (ESI-TOF) for C₁₂H₂₂N₂O₇: [M + H]⁺

calcd, 307.14998; found, 307.1494; [M + Na]⁺calcd, 329.13192;

found, 329.1312.

1,5-Bis-(2-acetamido-2-deoxy-r-D-glucopyranosyloxycar-

bonylamino)-3-oxa-pentane (21). According to GP 3, 16 (200 mg,

0.235 mmol) was dissolved in dry MeOH (5 mL). After addition

of a solution of NaOMe in MeOH (0.5 M, 275 µL, 138 µmol) and

stirring for 1.5 h, the reaction mixture was neutralized using acidic

ion-exchange resin (Amberlite IRC-86, H⁺form). Filtration and

crystallization gave 21 (140 mg) in quantitative yield as a white

solid. R_f) 0.19 (MeCN/H₂O 4:1); ¹H NMR (400.1 MHz, D₂O): δ

5.93 (d, J ) 2.9 Hz, 2H; H-1), 4.03 (dd, J ) 10.3, 3.3 Hz, 2H;

H-2), 3.81-3.70 (m, 8H; H-3, H-5, H-6a, H-6b), 3.62-3.53 (m,

6H; H-4, CH₂CH₂), 3.32 (br s, 4H; CH₂CH₂), 1.99 (s, 6H;

C(O)CH₃); ¹³C NMR (100.6 MHz, D₂O): δ 175.6 (C(O)CH₃), 157.5

(OC(O)NH), 92.6 (C-1), 74.8 (C-5), 71.7 (C-3), 70.4 (C-4), 70.1

(CH₂CH₂), 61.3 (C-6), 53.7 (C-2), 41.1 (CH₂CH₂), 22.8 (C(O)CH₃);

HR-MS (ESI-TOF) for C₂₂H₃₈N₄O₁₅: [M + H]⁺calcd, 599.24064;

found, 599.2381; [M + Na]⁺calcd, 621.22259; found, 621.2200.

1,4-Bis-(3-(2-acetamido-2-deoxy-r-D-glucopyranosyloxycar-

bonylamino)-propoxy)-butane (22). According to GP 3, 17 (141

mg, 148 µmol) was dissolved in dry MeOH (3 mL). After addition

of a solution of NaOMe in MeOH (0.54 M, 5 µL, 27 µmol) and

stirring for 2 h, the reaction mixture was neutralized using acidic

ion-exchange resin (Amberlite IRC-86, H⁺form). Filtration and

crystallization gave 22 (90 mg, 87%) as a white solid. R_f) 0.26

N,N′-Bis-(2-acetamido-3,4,6-tri-O-acetyl-2-deoxy-r-D-glu-

copyranosyloxycarbonyl)-4,7,10-trioxa-1,13-tridecandiamine

(18). According to GP 1, 5 (225 mg, 0.44 mmol), CH₂Cl₂(5 mL),

4,7,10-trioxa-1,13-tridecandiamine (44 µL, 0.2 mmol), and EtN(i-

Pr)₂(70 µL, 0.4 mmol) were reacted for 2 h. FC (CH₂Cl₂/MeOH

15:1) gave 18 (192 mg, 99%) as a white solid. R_f) 0.20 (CH₂Cl₂/

1

MeOH 15:1); H NMR (600.1 MHz, CDCl₃): δ 6.06 (d, J ) 3.6

1

Hz, 2H; H-1), 6.05 (br s, 2H; NHAc), 5.67 (t, J ) 5.6 Hz, 2H;

(MeCN/H₂O 4:1); H NMR (600.1 MHz, D₂O): δ 5.86 (br s, 2H;

9

8716 J. AM. CHEM. SOC. VOL. 132, NO. 25, 2010

Multivalent Binding to Wheat Germ Agglutinin

A R T I C L E S

H-1), 3.98 (dd, J ) 10.5, 2.1 Hz, 2H; H-2), 3.76-3.66 (m, 8H;

H-6a, H-6b, H-5, H-3), 3.52-3.46 (m, 10H; CH₂, CH₂, H-4), 3.16

(br s, 4H; CH₂), 1.96 (s, 6H; C(O)CH₃), 1.72 (t, J ) 6.0 Hz, 4H;

CH₂), 1.55 (br s, 4H, CH₂);¹³C NMR (150.9 MHz, D₂O): δ 175.4

(C(O)CH₃), 157.2 (OC(O)NH), 92.4 (C-1), 74.6 (C-5), 71.5 (C-3),

71.3 (CH₂), 70.2 (C-4), 68.5 (CH₂), 61.1 (C-6), 53.5 (C-2), 38.3

(CH₂), 29.5 (CH₂), 26.2 (CH₂), 22.6 (C(O)CH₃); HR-MS (ESI-TOF)

for C₂₈H₅₀N₄O₁₆: [M + H]⁺calcd, 699.32946; found, 699.3280;

[M + Na]⁺calcd, 721.3114; found, 721.3100.

N,N′-Bis-(2-acetamido-2-deoxy-r-D-glucopyranosyloxycar-

bonyl)-4,7,10-trioxa-1,13-tridecandiamine (23). According to GP

2, 18 (165 mg, 171 µmol) was dissolved in MeOH/EtMe₂N 5:1

and reacted for 2 days. Compound 23 was obtained in quantitative

yield as a white solid. R_f) 0.44 (MeCN/H₂O 2:1); ¹H NMR (600.1

MHz, D₂O): δ 5.86 (d, J ) 3.0 Hz, 2H; H-1), 3.50 (dd, J ) 10.7,

3.3 Hz, 2H; H-2), 3.76-3.66 (m, 8H; H-6a, H-6b, H-3, H-5), 3.61

(br s, 4H; OCH₂CH₂O), 3.60 (br s, 4H; OCH₂CH₂O), 3.52-3.48

(m, 6H; NCH₂CH₂CH₂O, H-4), 3.17-3.15 (m, 4H;

NCH₂CH₂CH₂O), 1.95 (s, 6H; C(O)CH₃), 1.74-1.72 (m, 4H;

NCH₂CH₂CH₂O);¹³C NMR (150.9 MHz, D₂O): δ 175.4 (C(O)CH₃),

157.2 (OC(O)NH), 92.4 (C-1), 74.7 (C-5), 71.5 (C-3), 70.4 (C-4),

70.2 (OCH₂CH₂O), 69.0 (NCH₂CH₂CH₂O), 61.6 (C-6), 53.5 (C-

2), 38.2 (NCH₂CH₂CH₂O), 29.5 (NCH₂CH₂CH₂O), 22.6

(C(O)CH₃); HR-MS (ESI-TOF) for C₂₈H₅₀N₄O₁₇: [M + H]⁺calcd,

715.32437; found, 715.3229; [M + Na]⁺calcd, 737.30632; found,

737.3046.

16 × 250 mm, ﬂow ) 9.6 mL min^-1, 1-10% MeCN in H₂O/0.1%

TFA over 20 min; t_R) 12.7 min) gave 28 (28 mg, 41% based on

initial resin loading). ¹H NMR (600.1 MHz, DMSO-d₆): δ

8.13-8.08 (1 H; NHAc D-Dab-1), 8.02-7.94 (1 H; NH D-Dab-2),

7.74-7.66 (2 H; NHAc GlcNAc), 7.33-7.29 (1 H; NH₂), 7.12-6.81

(3 H; 1 H from NH₂, 2 H from carbamate NH), 5.81-5.79 (2 H;

H-1 GlcNAc), 4.37-4.18 (2 H; C^RH from D-Dab-1 and D-Dab-2),

3.73-3.70 (2 H; H-2 GlcNAc), 3.58-3.43 (8 H + H₂O; H-6a,

H-6b, H-3, H-5 GlcNAc), 3.35-3.21 (2 H; H-4 GlcNAc), 3.08-2.92

(4 H; C^γH₂from D-Dab-1 and D-Dab-2), 1.89-1.81 (11 H; 9 H

from C(O)CH₃, C^ꢀH-a from D-Dab-1 and D-Dab-2), 1.69-1.58 (2

H; C^ꢀH-b from D-Dab-1 and D-Dab-2). ¹³C NMR (150.9 MHz,

DMSO-d₆): δ 172.6, 171.0 (C(O)CH), 169.3 (C(O)CH₃), 154.4

(OC(O)NH), 90.6 (C-1 GlcNAc), 74.4 (C-5 GlcNAc), 70.1 (C-3

GlcNAc), 70.0 (C-4 GlcNAc), 60.4 (C-6 GlcNAc), 52.8 (C-2

GlcNAc), 50.4 (C^RD-Dab-1), 50.1 (C^RD-Dab-2), 36.9 (C^γD-Dab-1

and D-Dab-2), 31.7 (C^ꢀD-Dab-1 and D-Dab-2), 22.4, 22.3

(C(O)CH₃). Analytical RP HPLC (Knauer Eurospher C18, 4 × 250

mm, ﬂow: 0.9 mL min^-1, 1-10% MeCN in H₂O/0.1% TFA over

20 min): t_R) 13.6 min. HR ESI-TOF-MS: calcd for C₂₈H₄₇N₇O₁₇,

754.31012 [M + H]⁺, 776.29206 [M + Na]⁺; found, 754.3080,

776.2896.

Enzyme-Linked Lectin Assay (ELLA). Assays were carried

out as previously described.²⁸Brieﬂy, microtiter plates with a

covalently immobilized ꢀ-GlcNAc derivative [4-(2-acetamido-2-

deoxy-ꢀ-D-glucopyranosyloxy)-(Z)-but-2-enyloxycarbonyl-(3-(2-(2-

(3-amino-propoxy)-ethoxy)-ethoxy)propyl)-amine] as reference ligand

were incubated with mixtures of horseradish peroxidase (HRP)-

labeled WGA (1 µg mL^-1) and the respective WGA ligand in

varying concentrations. After incubation, the plates were washed

and remaining labeled WGA bound to the reference ligand was

quantiﬁed by an HRP-catalyzed color reaction using 2,2′-azinobis(3-

ethylbenzothiazoline-6-sulfonic acid) diammonium salt (ABTS) as

substrate. From dose-response curves for inhibition of the binding

of HRP-labeled WGA to immobilized GlcNAc, the concentrations

that reduce the binding of labeled WGA to the microtiter plates by

50% (IC₅₀values) were determined as a measure of potency of the

synthesized inhibitors. All tests were performed in duplicate.

Puriﬁcation of WGA Isoforms.⁵⁰WGA was purchased from

Sigma as lyophilized powder. About 10 mg was dissolved in 1.5

mL of buffer A (100 mM Na acetate, 20 mM NaCl, pH 3.8). The

sample was applied with a ﬂow rate of 1 mL min^-1to a SP

Sepharose cation exchange column equilibrated with 3 column

volumes (CVs) of buffer A. The column was washed with 1 CV

of buffer A. WGA isoforms were eluted with 100 mL of a linear

gradient 0-100% buffer B (100 mM Na acetate, 500 mM NaCl,

pH 3.8). Fractions containing either WGA isoform 1, 2, or 3 were

pooled and concentrated with centrifugal ﬁlter devices (10 kDa

cutoff) to a ﬁnal concentration of approximately 8 mg mL^-1as

determined by measurement of E₂₈₀, using a theoretical extinction

coefﬁcient of 28 930 M^-1cm^-1, calculated based on amino acid

sequence.⁵¹

Tris-(2-(2-acetamido-2-deoxy-r-D-glucopyranosyloxycar-

bonylamino)-ethyl)-amine (24). According to GP 2, 19 (222 mg,

0.175 mmol) was dissolved in MeOH/EtMe₂N 5:1 and reacted for

5 days. Compound 24 was obtained in quantitative yield as a white

solid. R_f) 0.08 (MeCN/H₂O 4:1); ¹H NMR (600.1 MHz, D₂O): δ

5.88 (br s, 3H; H-1), 3.97-3.96 (m, 3H; H-2), 3.74-3.65 (m, 12H;

H-6a, H-6b, H-3, H-5), 3.50 (t, J ) 9.4 Hz, 3H; H-4), 3.17-3.16

(m, 6H; CH₂), 2.63 (br s, 6H; CH₂), 1.94 (s, 9H; C(O)CH₃);¹³C

NMR (150.9 MHz, D₂O): δ 175.2 (C(O)CH₃), 157.2 (OC(O)NH),

92.3 (C-1), 74.5 (C-5), 71.4 (C-3), 70.1 (C-4), 60.9 (C-6), 53.7

(CH₂), 53.4 (C-2), 39.0 (CH₂), 22.5 (C(O)CH₃); HR-MS (ESI-TOF)

for C₃₃H₅₇N₇O₂₁: [M + H]⁺calcd, 888.36803; found, 888.3651;

[M + Na]⁺calcd, 910.34997; found, 910.3470.

Ac-D-Dab(Ddv)-D-Dab(Ddv)-NH-Rink-PS (26). Solid-phase

peptide synthesis was carried out in 90 µmol scale in a syringe

reactor on Rink Amide polystyrene resin (Rapp Polymere, Tu¨bin-

gen, Germany) (abbreviation: Rink-PS, loading density 0.6 mmol

g^-1) following standard Fmoc protocols.³³Peptide couplings were

performed in NMP using Fmoc-D-Dab(Ddv)-OH (2 equiv), HBTU

(1.9 equiv), HOBt (3 equiv), and EtNi-Pr₂(4 equiv). The N-terminal

Fmoc group was removed with 20% piperidine in DMF (1 × 3

min, 1 × 15 min) and the resin was washed with DMF (10 × 1

min). The N-terminus was acylated with 10% Ac₂O in DMF (2 ×

15 min) followed by washing the resin with DMF (10 × 1 min).

Neoglycodipeptide (28). Resin 26 was deprotected by treatment

with hydrazine hydrate/DMF (4:96) (6 × 5 min) and washed with

DMF (10 × 1 min). After swelling for 10 min in EtNi-Pr₂(3 equiv)

in DMSO/DMF (1:3), the resin was treated with active carbonate

5 (4 equiv) and EtNi-Pr₂(4 equiv) in DMSO/DMF (1:3) overnight

and washed with DMF (10 × 1 min) and CH₂Cl₂. Cleavage from

the resin was achieved by treatment with TFA/i-Pr₃SiH/water (95:

2.5:2.5) (1 × 3 h, 1 × 1 h). The cleavage solutions were dropped

into ice-cold tert-butylmethylether (10 times volume) and stored

at -18 °C overnight. Precipitated product was collected by

centrifugation at 4 °C and 4500 rpm for 30 min. The resulting pellet

was resuspended in tert-butylmethylether and the centrifugation

procedure was repeated. The pellet was then lyophilized from water

to yield crude acetylated neoglycodipeptide 27. MALDI-MS: for

C₄₀H₅₉N₇O₂₃calcd, 1028.4 [M + Na]⁺, 1044.3 [M + K]⁺; found,

1028.3, 1044.3. Acetylated 27 was deprotected by treatment with

NaOMe (27 mM) in dry MeOH/CHCl₃(1:1) for 3.5 h. The solution

was neutralized with 30% acetic acid and the solvent was removed

under vacuum. Puriﬁcation by RP-HPLC (Knauer Eurospher C18,

Cocrystals of WGA1 with GlcNAc-Carbamate 20. During

protein concentration, buffer was exchanged against crystallization

buffer (20 mM Na Acetate, 6 mM CaCl₂, 4% Ethanol at pH 4.9).

Small volumes of a solution of 20 in crystallization buffer (0.1 M)

were added to the protein sample at 8 mg mL^-1in order to reach

a ﬁnal ligand concentration of 7 mM. The samples were thoroughly

vortexed and incubated on ice for 1 h. Crystallization was performed

in hanging drop vapor diffusion setups using 1 mL reservoir solution

R (0.04 M KH₂PO₄, 20% glycerol, 14-17% PEG 8000) and a drop

size of 6 µL (4 µL protein solution + 2 µL reservoir solution). At

18 °C, cocrystals appeared overnight with a size of up to 0.3 mm.

Cocrystals of WGA3 with Divalent Ligand 23. During protein

concentration, buffer was exchanged against buffer B. Small

(50) Rice, R. H.; Etzler, M. E. Biochemistry 1975, 14, 4093–4099.

(51) Gasteiger, E.; Hoogland, C.; Gattiker, A.; Duvaud, S.; Wilkins, M. R.;

Appel, R. D.; Bairoch, A. In The Proteomics Protocols Handbook;

Walker, J. M., Ed.; Humana Press: Totowa, 2005; pp 571-607.

9

J. AM. CHEM. SOC. VOL. 132, NO. 25, 2010 8717

A R T I C L E S

Schwefel et al.

volumes of a solution of 23 in crystallization buffer (0.1 M) were

added to the protein sample at 8 mg mL^-1in order to reach a ﬁnal

ligand concentration of 1 mM. The samples were thoroughly

vortexed and incubated on ice for 1 h. Crystallization was performed

with hanging drop vapor diffusion setups using 1 mL of reservoir

solution R and a drop size of 6 µL (4 µL protein solution + 2 µL

reservoir solution). At 18 °C, cocrystals appeared after 3 weeks

with a size of up to 0.5 mm in the largest dimension.

ligand 2 was reﬁned with PHENIX⁶⁰applying restrained positional

reﬁnement and TLS reﬁnement. Five percent of recorded reﬂections

were randomly set aside for cross validation. Restraints for the

visible part of the cyclic peptide ligand 2 were generated using

Coot and the CCP4i monomer sketcher based on the energy-

minimized NMR structure with least restraint violations. Electron

density map inspection and model building were done using Coot.

Geometry was validated and Ramachandran plot was calculated

using the Molprobity server.⁶¹Distances were measured using Coot.

Hydrogen bonding was analyzed using UCSF Chimera.⁶²Figures

containing protein X-ray structure representations were prepared

using PyMol.⁶³All molecular surface representations are solvent-

excluded (Connolly) surfaces, calculated with a probe radius of

1.4 Å.

NMR Spectroscopy. NMR samples contained cyclic neoglyco-

peptide 2 (3-5 mM), 0.05% NaN₃, and 3-(trimethylsilyl)-propionic

acid-2,2,3,3-d₄(TSP) (0.5 mM) in H₂O/D₂O (95:5). The pH was

adjusted to 5.0 with aqueous hydrochloric acid and sodium

hydroxide solution. Final concentrations of sodium chloride were

in the low millimolar range. For experiments in D₂O, the sample

was ﬁrst dried by evaporation and coevaporated with D₂O to ensure

quantitative isotopic exchange. The residual dry substance was

redissolved in D₂O. Before NMR experiments were performed, the

pH was readjusted to 5.0 with NaOD. NMR spectra were recorded

at 300 K on a Bruker Avance DRX 600 spectrometer equipped

with an inverse 5 mm TXI-H/C/N triple resonance probe with

shielded z-gradients. Assignments and calculation of distance and

torsion restraints were carried out with the help of standard 2D

experiments that were detected according to the States-TPPI method

in a phase sensitive mode. 2D spectra were typically recorded with

4k data points in F2 (1k to 1.5k data points in ¹³C-HSQC spectra)

and 256 or 512 experiments in F1, with 16-64 transients per

increment. The TOCSY experiments were performed with a

spinlock ﬁeld of 10.0 kHz using the MLEV17 sequence with a

mixing time of 100 ms. The WATERGATE scheme⁶⁴was used to

suppress the H₂O resonances in ¹H, TOCSY and NOESY spectra.

In 2D COSY spectra, H₂O resonances were suppressed by pre-

saturation of the H₂O resonances. The spectra were processed with

simple zero ﬁlling in both dimensions and referenced on TSP using

Topspin 1.3.b.20. Cara 1.5.3 was used in order to extract distance

restraints from NOESY spectra that were detected with mixing times

of 50 ms. Mestrec 4.6.2.3 and Topspin 1.3.b.20 were used to extract

coupling constants from DQF COSY spectra. These were translated

into dihedral angle restraints with the aid of a Karplus relationship

(³J_HH) A cos²φ + B cos φ + C, parameters: A ) 7.9, B ) -1.55,

C ) 1.35).⁶⁵

Crystals of WGA1, Cross-Linking, and Soaking with

Tetravalent Cyclic Peptide Ligand 2. During protein concentration

of WGA1, buffer was exchanged against buffer B. Hanging drop

vapor diffusion experiments were set up using 1 mL of reservoir

solution R and a drop size of 6 µL (4 µL protein solution + 2 µL

reservoir solution). At 8 °C, crystals of WGA1 appeared overnight

and had a size of approximately 0.1 mm. Soaking of WGA1 crystals

with tetravalent ligand 2 was unsuccessful because the crystals

cracked and dissolved upon addition of the compound. We therefore

employed glutaraldehyde cross-linking. Crystals of WGA1 in

complex with GlcNAc⁴²were chosen, because of the large solvent

channels present in this crystal form and because native WGA1

crystals were reproducibly twinned. The crystals were extensively

washed in stabilizing solution (0.04 M KH₂PO₄, 20% (w/v) glycerol,

20% (w/v) PEG 8000) to remove GlcNAc. Cross-linking was

performed using 25% glutaraldehyde in a microbridge,⁵²with the

exception that cross-linking was prolonged to 12 h at 4 °C.

Afterward, crystals were washed again in stabilizing solution and

soaked by transfer to stabilizing solution containing approximately

0.2 mM cyclic peptide ligand 2. Crystals were soaked for 24 h at

4 °C prior to freezing.

Data Collection and Crystal Structure Analysis. Crystals were

mounted in cryogenic loops (Hampton) and frozen by plunging into

liquid nitrogen. Diffraction data were recorded using an ADSC

Q315r detector at beamline ID23-1, ESRF, Grenoble or using a

MARresearch MX-225 CCD detector at beamline X06SA, SLS,

Villigen. In all cases, data were collected at 100 K. A total of 360

frames per crystal were recorded with an oscillation range of 0.5°

and an exposure time of 0.5 or 1 s per frame. X-ray wavelengths

are listed in Table 2. For all data sets, data reduction was carried

out using the program XDS.⁵³Data collection statistics are listed

in Table 2.

Structures were solved by molecular replacement with known

WGA atomic coordinates, using the program Molrep.⁵⁴As starting

model, native WGA1 coordinates (PDB ID: 7WGA²⁴) or native

WGA3 coordinates (PDB ID: 1WGT^23c) were chosen. Ligands were

only included in the model when unbiased positive F_o- F_cdensity

was visible. The structures of WGA1 bound to compound 20 and

WGA3 in complex with divalent ligand 23 were reﬁned using the

program Refmac.⁵⁵The WGA3∼23 crystal was merohedrally

twinned with the twin operator -h-k, k, -l and a twin fraction of

0.44, as detected by SFCHECK.⁵⁶Ten percent of all recorded

reﬂections were excluded for cross validation taking into account

the twin operator to ensure that twin-related reﬂections are in the

same set. Ligand restraints were generated using the CCP4i

monomer sketcher,⁵⁷Coot,⁵⁸and the PRODRG server.⁵⁹The

structure of WGA1 complexed with the tetravalent cyclic peptide

Structure Calculation. All molecular dynamics (MD) simula-

tions were performed using sander of the AMBER 8 package on a

computer cluster (Transtec, 8 Opteron Dual-Core processors, 32

GB RAM, CentOS Linux). The simulations were done by com-

bining parameters for the peptide backbone from the PARM94/

PARM99 force ﬁelds developed by Kollman et al.⁶⁶with those

developed by Woods and co-workers for carbohydrate structures.⁶⁷

As the structure calculations required AMBER-compatible topo-

logical descriptions of all contained structural fragments, topology

(60) Adams, P. D.; Grosse-Kunstleve, R. W.; Hung, L. W.; Ioerger, T. R.;

McCoy, A. J.; Moriarty, N. W.; Read, R. J.; Sacchettini, J. C.; Sauter,

N. K.; Terwilliger, T. C. Acta Crystallogr., Sect. D: Biol. Crystallogr.

2002, 58, 1948–1954.

(61) Lovell, S. C.; Davis, I. W.; Arendall, W. B., III; de Bakker, P. I. W.;

Word, J. M.; Prisant, M. G.; Richardson, J. S.; Richardson, D. C.

Proteins: Struct., Funct., Genet. 2003, 50, 437–450.

(62) Pettersen, E. F.; Goddard, T. D.; Huang, C. C.; Couch, G. S.;

Greenblatt, D. M.; Meng, E. C.; Ferrin, T. E. J. Comput. Chem. 2004,

25, 1605–1612.

(63) DeLano, W. L. The PyMOL Molecular Graphics System; DeLano

Scientiﬁc: Palo Alto, CA, 2002.

(64) Piotto, M.; Saudek, V.; Sklenar, V. J. Biomol. NMR 1992, 2, 661–

665.

(65) Ramachandran, G. N.; Chandrasekaran, R.; Kopple, K. D. Biopolymers

1971, 10, 2113–2131.

(52) Lusty, C. J. J. Appl. Crystallogr. 1999, 32, 106–112.

(53) Kabsch, W. J. Appl. Crystallogr. 1993, 26, 795–800.

(54) Vagin, A.; Teplyakov, A. J. Appl. Crystallogr. 1997, 30, 1022–1025.

(55) Murshudov, G. N.; Vagin, A. A.; Dodson, E. J. Acta Crystallogr.,

Sect. D: Biol. Crystallogr. 1997, 53, 240–255.

(56) Vaguine, A. A.; Richelle, J.; Wodak, S. J. Acta Crystallogr., Sect. D:

Biol. Crystallogr. 1999, 55, 191–205.

(57) Potterton, E.; Briggs, P.; Turkenburg, M.; Dodson, E. Acta Crystallogr.,

Sect. D: Biol. Crystallogr. 2003, 59, 1131–1137.

(58) Emsley, P.; Cowtan, K. Acta Crystallogr., Sect. D: Biol. Crystallogr.

2004, 60, 2126–2132.

(59) Schuettelkopf, A. W.; van Aalten, D. M. F. Acta Crystallogr., Sect.

D: Biol. Crystallogr. 2004, 60, 1355–1363.

9

8718 J. AM. CHEM. SOC. VOL. 132, NO. 25, 2010

Multivalent Binding to Wheat Germ Agglutinin

A R T I C L E S

ﬁles were generated accordingly for several artiﬁcial residues

contained in the neoglycopeptide. For Lys and Glu residues with

side chains involved in an amide bond, for D-Dab, and for D-Val

residues, topology ﬁles were created in the format of the amino

acid library in use. AMBER-compatible topological descriptions

of the GlcNAc- the Boc, and the ꢀ-Ala residues were generated

with the programs antechamber and prepgen that are part of

AMBER 8. Partial charges were used as given in the library

all_amino94.lib or calculated using prepgen, that are both part of

the AMBER 8 package. For MD simulations in vacuum, these

charges were reduced to 20% of their original values.

Acknowledgment. This Article is dedicated to Professor Horst

Kessler on the occasion of his 70th birthday. Financial support by

the Deutsche Forschungsgemeinschaft (FOR 434 “Oligosaccharide

and DNA Chips - Analysis of Secondary Gene Products”), the

University of Konstanz, the Konstanz Research School Chemical

Biology, and the BMBF is gratefully acknowledged. J.G.B. is

grateful for a stipend of the Friedrich-Ebert-Stiftung. We thank

Konrad Bergen and Philippe Reuter for their help with crystalliza-

tion and the staff at the SLS and the ESRF for assistance in crystal

data collection.

A structurally diverse ensemble of 100 conformers of neogly-

copeptide 2 was generated by 150 ps unconstrained molecular

dynamics simulation at 2000 K. Subsequently, the ensemble was

subjected to the following steps in the presence of NMR constraints:

Simulated annealing succeeded by energy-minimization in vacuum

followed by simulated annealing and subsequent energy minimiza-

tion employing a generalized Born solvation model.⁶⁸

Supporting Information Available: Inhibition curves, electron

density plots, additional data and ﬁgures on structure determi-

nation of 2, and NMR spectra of 5, 15-24, and 28. This material

is available free of charge via the Internet at http://pubs.acs.org.

JA101646K

(66) (a) Cornell, W. D.; Cieplak, P.; Bayly, C. I.; Gould, I. R.; Merz, K. M.,

Jr.; Ferguson, D. M.; Spellmeyer, D. C.; Fox, T.; Caldwell, J. W.;

Kollman, P. A. J. Am. Chem. Soc. 1995, 117, 5179–5197. (b)

Cheatham, T. E., 3rd; Cieplak, P.; Kollman, P. A. J. Biomol. Struct.

Dyn. 1999, 16, 845–862.

(67) Woods, R. J.; Dwek, R. A.; Edge, C. J.; Fraser-Reid, B. J. Phys. Chem.

1995, 99, 3832–3846.

(68) Xia, B.; Tsui, V.; Case, D. A.; Dyson, H. J.; Wright, P. E. J. Biomol.

NMR 2002, 22, 317–331.

9

J. AM. CHEM. SOC. VOL. 132, NO. 25, 2010 8719

Article Doi

Structural basis of multivalent binding to wheat germ agglutinin

DOI: 10.1021/ja101646k

Source and publish data:

Authors:

Article abstract of DOI:10.1021/ja101646k

Full text of DOI:10.1021/ja101646k

Products guided by the article

R&D Labs maybe for 1233921-11-3

Relevant to this article

Hot Product