PROTEINS: Structure, Function, and Genetics 28:268–284 (1997) Role of Aromatic Amino Acids in Carbohydrate Binding of Plant Lectins: Laser Photo Chemically Induced Dynamic Nuclear Polarization Study of Hevein Domain-Containing Lectins Hans-Christian Siebert,1,2 Claus-Wilhelm von der Lieth,3 Robert Kaptein,2 Jaap J. Beintema,4 Klaas Dijkstra,4 Nico van Nuland,4 Ukun M. S. Soedjanaatmadja,5 Ann Rice,6 Johannes F. G. Vliegenthart,2 Christine S. Wright,6 and Hans-Joachim Gabius1* 1Institut für Physiologische Chemie, Tierärztliche Fakultät, Ludwig-Maximilians-Universität, München, Germany 2Bijvoet Center for Biomolecular Research, University of Utrecht, Utrecht, The Netherlands 3Zentrale Spektroskopie, Deutsches Krebsforschungszentrum, Heidelberg, Germany 4Biochemisch Laboratorium and BIOSON Research Institute, Department of Biophysical Chemistry, University of Groningen, Groningen, The Netherlands 5Laboratorium Biokimia, Jl. Singaperbangsa 2, Padjadjaran University, Bandung, Indonesia 6Department of Biochemistry and Molecular Biophysics, Virginia Commonwealth University, Richmond, Virginia ABSTRACT Carbohydrate recognition by lectins often involves the side chains of tyrosine, tryptophan, and histidine residues. These moieties are able to produce chemically induced dynamic nuclear polarization (CIDNP) signals after laser irradiation in the presence of a suitable radical pair-generating dye. Elicitation of such a response in proteins implies accessibility of the respective groups to the light-absorbing dye. In principle, this technique is suitable to monitor surface properties of a receptor and the effect of ligand binding if CIDNP-reactive amino acids are affected. The application of this method in glycosciences can provide insights into the protein-carbohydrate interaction process, as illustrated in this initial study. It focuses on a series of N-acetylglucosamine-binding plant lectins of increasing structural complexity (hevein, pseudohevein, Urtica dioica agglutinin and wheat germ agglutinin and its domain B), for which structural NMR- or X-ray crystallographic data permit a decision of the validity of the CIDNP methodderived conclusions. On the other hand, the CIDNP data presented in this study can be used for a rating of our molecular models of hevein, pseudohevein, and domain B obtained by various modeling techniques. Experimentally, the shape and intensity of CIDNP signals are determined in the absence and in the presence of specific glycoligands. When the carbohydrate ligand is bound, CIDNP signals of side chain protons of tyrosine, tryptophan, or histidine residues are altered, for example, they are broadened and of reduced intensity or disappear completely. In the case of UDA, the appearance of a new tryptophan signal upon ligand binding was interpreted as an indication for a r 1997 WILEY-LISS, INC. conformational change of the corresponding indole ring. Therefore, CIDNP represents a suitable tool to study protein-carbohydrate interactions in solution, complementing methods such as X-ray crystallography, high-resolution multidimensional nuclear magnetic resonance, transferred nuclear Overhauser effect experiments, and molecular modeling. Proteins 28:268–284, 1997 r 1997 Wiley-Liss, Inc. Key words: lectins; agglutinins; chemically induced dynamic nuclear polarization (CIDNP); nuclear magnetic resonance (NMR); molecular modeling INTRODUCTION The recognition of carbohydrate determinants by lectins plays an important role in mediation of intercellular binding and elicitation of biosignaling processes.1–4 Structural aspects of this glycobiological interplay can, in principle, be visualized by multidimensional nuclear magnetic resonance (NMR) experiments and/or X-ray crystallography. However, these techniques can suffer from a number of inherent restrictions, as, for example, molecular mass limitations in multidimensional NMR studies and in X-ray crystallography a lack of information about the dynamic behavior of molecules in solution. There- Contract grant sponsor: Human Capital and Mobility Program of the European Community *Correspondence to: Hans-Joachim Gabius, Institut für Physiologische Chemie, Tierärztliche Fakultät, Ludwig-MaximiliansUniversität, 80539 München, Germany. Received 9 September 1996; Accepted 3 January 1997 CARBOHYDRATE RECOGNITION BY LECTINS fore, additional methods to reliably assess defined structural aspects deserve attention. CIDNP experiments, which allow the detection of the amino acid side chains of tyrosine, tryptophan, and histidine at protein surfaces,5 offer such a possibility. They readily provide information on their targets following changes of external parameters such as temperature and/or pH values, as demonstrated in studies of several biomolecules, and they are not very timeconsuming.6,7 The CIDNP technique has previously been used for comparative studies of nonspecific and specific interactions between the lac-repressor headpiece and DNA,8,9 denatured states of lysozyme10 as well as of glycoproteins in glycosylated and deglycosylated form11 or in sialylated and desialylated form12 in solution. So far, the CIDNP method has not been applied to a carbohydrate-binding protein, although it is known that side chains of tyrosine, tryptophan, and histidine can play an essential role in ligand binding of carbohydrate receptors apart from their well-appreciated participation in stabilizing protein conformation.13–18 Since only three types of amino acids are under consideration for these measurements in contrast to conventional 1H-NMR experiments, the problem of overlapping signals is clearly less stringent. When structural parameters from the nuclear Overhauser effect (NOE)-dependent NMR assessment or from X-ray crystallography are available, the calculation of surface accessibilities for the respective types of side chains enables a useful comparison of these datasets with the CIDNP results. To determine the actual value of this method in the field of glycosciences, several well-characterized Nacetylglucosamine-binding lectins from different species with increasing molecular complexity are selected as models. In detail, laser photo CIDNP experiments of hevein, pseudohevein, Urtica dioica agglutinin (UDA), wheat germ agglutinin (WGA), and the B domain (domB) of WGA1 have been carried out in the absence and in the presence of their specific ligands. Hevein (43 amino acid residues) and pseudohevein (43 amino acid residues) are small lectins of rubber trees (Hevea brasiliensis) with ragged C-terminal sequences.19,20 In addition to their assumed role in latex coagulation they are considered to serve a protective function similar to that of the related lectins WGA and UDA.21–24 The ratio of expression of hevein to pseudohevein is 10:1, and the amino acid sequences of the two proteins differ only in six positions from each other.19 Among these differences the replacement of Trp21, present in hevein, by a tyrosine residue in pseudohevein, is especially notable in the context of the CIDNP experiments. Another member of the chitin-binding family of plant lectins is UDA from stinging nettle, whose structure consists of two heveinlike domains. It has a single chain of 89 amino acids, comprising two binding sites with different affinities for oligom- 269 Fig. 1. Laser photo CIDNP difference spectra (aromatic part) of hevein and hevein–(GlcNAc)2 complexes at pD 4. a: Ligandfree hevein. b: 1 mmol hevein 1 1 mmol (GlcNAc)2. 270 H.-C. SIEBERT ET AL. TABLE I. Surface Accessibilities [Å2] of CIDNP-Reactive Amino Acid Groups in the 8 Hevein Structures That Represent Relative Local Energy Minima According to MD Simulations* I II III IV V VI VII VIII x 160.9 87.6 32.5 75.0 148.7 71.1 20.1 57.0 162.8 76.1 19.6 92.6 143.0 101.8 22.7 94.4 167.9 72.3 32.0 106.9 121.1 90.8 23.0 111.1 151.4 83.7 25.3 89.9 125.9 75.4 28.2 55.3 125.2 70.2 3.2 50.9 137.4 70.4 0.0 85.3 144.6 103.0 0.0 88.2 149.7 82.4 26.6 98.4 113.4 98.6 0.0 105.4 133.9 79.8 11.5 78.1 a. Dot density: 1, sphere radius: 1.5 Å Trp21 Trp23 Tyr30 His35 131.8 82.0 24.8 93.0 174.6 87.9 27.6 88.8 b. Dot density: 1, sphere radius: 4 Å Trp21 Trp23 Tyr30 His35 122.1 68.4 17.8 65.6 153.1 69.8 15.9 75.5 *MD simulations were carried out in the GROMOS (Groningen molecular simulation) force field.44,45 TABLE II. Surface Accessibilities [Å2] of CIDNP-Reactive Amino Acid Groups in 7 NMR- and MD-Derived Hevein Structures From the Protein Data Bank I II III IV V VI† VII† x 120.7 88.0 25.0 83.8 123.9 88.1 56.1 70.4 134.4 91.0 58.9 97.3 121.9 96.3 56.3 44.2 120.5 88.4 49.6 35.4 130.4 89.6 51.1 72.7 125.5 90.0 44.8 61.3 118.5 95.2 49.3 53.7 110.8 90.6 44.7 59.5 117.7 98.8 44.4 36.3 114.3 94.7 41.6 29.3 118.5 89.9 43.3 51.3 a. Dot density: 1, sphere radius: 1.5 Å Trp21 Trp23 Tyr30 His35 150.4 91.5 46.9 81.6 141.1 83.6 65.0 96.1 b. Dot density: 1, sphere radius: 4 Å Trp21 Trp23 Tyr30 His35 127.7 77.6 31.3 59.6 115.2 82.1 47.2 59.2 *MD simulations were carried out using the CHARMM (chemistry at Harvard macromolecular mechanics) force field.66 VI and VII are newly refined NMR structures of hevein and the respective information was kindly provided by Dr. N. H. Andersen and Dr. B. Cao. †Structures TABLE III. Surface Accessibilities [Å2] of CIDNPReactive Amino Acid Groups in 5 Hevein Structures That Represent Relative Local Energy Minima According to MD Simulations* Trp21 Trp23 Tyr30 His35 I II III IV V x 113.1 108.3 63.9 99.0 106.9 99.6 53.8 83.1 113.5 104.1 55.5 107.5 110.5 97.2 70.1 107.3 116.0 109.7 75.2 114.9 112.0 103.8 63.7 102.4 *Dot density: 1, sphere radius: 1.5 Å. MD simulations were carried out in the CVFF (consistence valence force field).42,43 ers of N-acetylglucosamine (GlcNAc).25,26 Besides its putative role in defense mechanisms in the plant, it is interesting to note that it triggers immunomodulatory responses in mammalian cells such as carbohydrate-dependent cytokine release or T-cell subset mitogenicity.27 These activities point to a potential perspective for clinical application, as is currently under investigation for the mistletoe lectin.28–30 The isoforms of dimeric WGA, which originate from hexaploid wheat (Triticum aestivum), also bind GlcNAc oligomers. Each chain of 17 kDa is divided into four domains with sequence repeats of 43 amino acids in length, suggesting evolution by gene duplication and fusion.31–33 Their structures display increased structural complexity in comparison to hevein and UDA. In contrast to hevein, pseudohevein, and UDA, WGA exhibits also a binding specificity for N-acetylneuraminic acid (NeuNAc). Moreover, the carbohydrate-binding sites of WGA are composed of regions contributed by domains on different subunits that are in contact with the subunit/subunit interface. Occupancy by either NeuNAc or GlcNAc oligosaccharides has been observed in crystal structures at all these sites.14,34–37 With the help of recombinant techniques it is possible to prepare single domains of the lectin for assessment of their individual binding properties. As the preferentially occupied site, domain B has been chosen for this purpose.13,38 In this study, the single domain of the more complex lectin WGA1 with its four covalently linked units is tested to compare its behavior to that of the single-domain proteins hevein as well as pseudohevein. These data for one domain derived from WGA can also contrib- CARBOHYDRATE RECOGNITION BY LECTINS 271 Fig. 2. Laser photo CIDNP difference spectra (aromatic part) of hevein and hevein–(GlcNAc)4 complexes at pD 3. a: Ligand-free hevein. b: 1 mmol hevein 1 1 mmol (GlcNAc)4. c: 1 mmol hevein 1 2 mmol (GlcNAc)4. TABLE IV. Comparison of Surface Accessibilities [Å2] of CIDNP-Reactive Amino Acid Groups in Hevein Structures Obtained Either by NMR Measurements in the Presence of the Ligand or by Computer-Assisted Calculations of Hevein in the Absence of the Ligand on the Basis of These NMR Data* Trp21 Trp23 Tyr30 His35 C SD H SD D 93.0 71.3 27.0 71.3 7.1 7.9 5.6 9.1 122.5 96.6 60.1 68.8 8.1 7.0 7.2 9.0 229.5 225.3 233.1 12.5 *Dot density: 1, sphere radius: 1.5 Å. C, hevein-ligand complex; H, ligand-free hevein; SD, standard deviation. ute to simplify the interpretation of the four hevein domain-containing hololectin. MATERIALS AND METHODS Carbohydrate Ligands N,N8-Diacetylchitobiose, N,N8,N9-triacetylchitotriose, N,N8,N9,N--tetraacetylchitotetraose, N-acetylneuraminic acid(a2-3)lactose, N-acetylneuraminic acid(a2-6)lactose and 2-deoxy-2,3-didehydro-Nacetylneuraminic acid (Neu5Ac2en) were purchased from Sigma (Munich, Germany). Deuterated water (99.96%) was obtained from Merck, Sharp and Dohme (Montreal, Canada). Flavin I mononucleotide was a gift from Dr. F. Müller (Sandoz, Basel). All other reagents were of analytical grade. Carbohydrate-Binding Proteins Hevein and pseudohevein were isolated from the latex of Hevea brasiliensis, as described elsewhere.19 Rubber latex of clones GT.1, PR-261 and LCB-1320 of Hevea brasiliensis were collected in several plantations (PTP XII Jalupang and PTP XIII Wangun Reja) in the area of Subang (Western Java, Indonesia), cooled on ice and centrifuged at high speed to collect the lutoid body fraction. Repeated freezing and thawing of this fraction followed by centrifugation yielded B serum, which was suspended at 4°C in water (4 mg/ml) after lyophilization and processed according to established procedures to supply hevein and pseudohevein. 272 H.-C. SIEBERT ET AL. Fig. 3. Laser photo CIDNP difference spectra (aromatic part) of pseudohevein and pseudohevein–(GlcNAc)4 complexes. a: Ligand-free pseudohevein. b: 1 mmol pseudohevein 1 1 mmol (GlcNAc)4. c: Signal intensity of Trp23e3 in dependence of the ligand concentration. TABLE V. Surface Accessibilities [Å2] of CIDNP-Reactive Amino Acid Groups of 10 Pseudohevein NMR Structures That Represent Relative Local Energy Minima According to MD Simulations* Tyr21 Trp23 Tyr30 I II III IV V VI VII VIII IX X x 82.9 123.7 30.1 91.9 119.8 27.9 94.1 124.8 24.7 83.2 126.5 27.1 84.7 125.1 36.3 87.9 130.8 39.3 88.3 143.4 36.0 89.8 114.8 37.2 84.3 118.3 36.7 103.4 116.2 36.9 89.1 127.3 33.2 *Dot density: 1, sphere radius: 1.5 Å. MD simulations were carried out in the GROMOS (Groningen molecular simulation) force field.44,45 The lectin from commercially available rhizomes of Urtica dioica was purified by successive steps, employing affinity chromatography on chitin, baseand acid-washed to remove an agglutinating substance that is associated with chitin, and subjected to cation exchange chromatography on sulfopropylSephadex C-50. Experimental details have been given elsewhere.26,39 Additionally, gel filtration on Biogel P-10 was performed to remove any aggregates. The lectin was judged to be homogeneous by SDS-PAGE analysis. WGA was isolated from raw wheat germ, as described previously.40 Clones of the 43-residue domain B of WGA were generated by using a fusion protein construct for convenient purification and detection.38 Briefly, two overlapping synthetic oligonucleotides were de- signed which encompass the complete coding region of domain B, including cleavage sites for restriction enzymes at each end, and the heveinlike terminal sequence G E Q. These were amplified by the polymerase chain reaction and ligated into the plasmid pUC 18, previously engineered to encode a truncated sequence of hepatitis B antigen followed by an arginine moiety for tryptic cleavage. The core-domB fusion product was expressed in Escherichia coli and purified by hydroxyapatite and Sepharose CL-4B column chromatography in successive steps according to a described method.41 The fusion partners were separated by trypsin digestion in Tris-buffered saline, pH 8.0. Optimal yields of domB were obtained in 2M guanidine-HCl at protein concentrations ,2.5 mg/ml. domB was isolated from the supernatant of this digest after precipitation of core by either lower- 273 CARBOHYDRATE RECOGNITION BY LECTINS TABLE VI. Surface Accessibilities [Å2] of CIDNP-Reactive Amino Acid Groups of 10 Pseudohevein NMR Structures That Represent Relative Local Energy Minima According to MD Simulations* I II III IV V VI VII VIII IX X x 82.2 132.4 44.1 78.8 122.2 30.7 79.0 131.9 40.7 129.0 100.8 83.8 133.3 128.9 81.2 122.4 106.3 89.0 a. Start structure: hevein structure, manual amino acid replacement. Dot density: 1, sphere radius: 1.5 Å Tyr21 Trp23 Tyr30 86.1 122.9 43.6 82.4 122.6 47.6 83.2 134.8 33.5 74.7 132.3 35.8 73.5 141.9 36.9 68.2 139.7 43.7 81.8 137.7 44.5 78.7 132.8 46.3 b. Start structure: ProMod based on hevein structure. Dot density: 1, sphere radius: 1.5 Å Tyr21 Trp23 Tyr30 107.2 96.9 73.3 111.8 110.7 79.7 104.4 104.5 90.3 113.8 106.5 72.1 133.3 104.5 95.1 119.2 102.8 110.8 125.7 107.1 103.3 146.3 100.2 100.8 *MD simulations were carried out in the CVFF (consistence valence force field).42,43 Fig. 4. Energy minimum conformation of pseudohevein, emphasizing the surface exposition of Tyr21 and Trp23 in the upper part of the picture. ing the pH to 4.5 or with 45% ammonium sulfate. The purity and oligomeric state of domB were determined by using HPLC, P-30 gel filtration, amino acid sequence analysis, and mass spectrometry. The ability of domB to bind GlcNAcN oligomers was verified by chitin affinity chromatography, tryptophan fluorescence spectroscopy, and isothermal titration calorimetry. CIDNP Method CIDNP experiments were performed at 360 MHz on a Bruker AM-360 NMR spectrometer. The light of a continuous-wave argon ion laser (Spectra Physics, Mountain View, USA) that operates in the multiline mode with principal wavelengths of 488.0 and 514.5 nm was directed to the sample by an optical fibre and 274 H.-C. SIEBERT ET AL. By alternately recording a light and four dark freeinduction decays, Fourier transformation and subtraction of the resulting light spectrum from the dark spectrum, the CIDNP difference spectrum was established, containing only lines of polarized residues. The tyrosine-dependent CIDNP effect corresponds to a spin-density distribution of the intermediate phenoxy radical with strong negative signals of the e1, e2 protons and less intense positive signals for the d1, d2 protons. The CIDNP spectrum of tryptophan is caused by an intermediate radical with strong spin density at the d1, e3, and h2 positions of the indole unit and very small spin density at the z2 and z3 positions, which all lead to positive CIDNP signals. CIDNP difference spectra of histidine show positive singlet signals for proton e1 and proton d2.6 Two-Dimensional NMR Experiments Fig. 5. Laser photo CIDNP difference spectra (aromatic part) of domB (WGA1) and domB (WGA1)–(GlcNAc)4 complexes. a: 0.2 mmol ligand-free domB (WGA1). b: 0.2 mmol domB (WGA1) 1 0.5 mmol (GlcNAc)4. chopped by a mechanical shutter controlled by the spectrometer. Typical operation conditions were 1 s presaturation pulse for water suppression, 0.5second light pulse (5 W), 5-µs RF pulse (90° flip angle), 1-second acquisition time, and 5-second delay. The duration of the light pulse is short enough to prevent serious heating due to light absorption. 8, 16, or 32 light scans gave an adequate signal-tonoise ratio for the tested samples. CIDNP was generated by using flavin I mononucleotide. Deuterated water was used as solvent for the assay mixture containing 0.2 to 2 mmol lectin, 0.5 to 2 mmol ligand, and 0.4 mmol flavin. The pD was calculated by adding 0.4 to the reading of a conventional pH meter, equipped with a glass electrode. Chemical shifts were determined relative to the position of the acetone signal (2.225 ppm). The broad signal in the region 4.7–4.8 ppm arose from the partially suppressed resonance of residual HOD (signal of partly deuterated water) at a ppm value of 4.76 ppm (pD 5 5.8, T 5 300 K). Briefly, the individual measurements were carried out in the following order of steps, as outlined in detail elsewhere.6 After the addition of a flavin dye to the protein solution, the sample was irradiated by an argon-ion laser during the light part of the CIDNP experiment. The photoexcited dye reacts reversibly with surface-exposed amino acid side chains of histidine, tyrosine, and tryptophan, thereby generating protein-dye radical pairs. Nuclear spin-polarization was obtained from backreactions of the radical pairs. To enable proton assignment and independent verification of the apparent consequences of the CIDNP experiments, TOCSY (total correlated spectroscopy) and NOESY (nuclear Overhauser effect spectroscopy) spectra of hevein, pseudohevein, and domB were recorded in H2O as well as in D2O at temperatures of 293 or 300 K with mixing times from 100 to 200 ms at 500 or 600 MHz on a Bruker 500 or 600 AMX-NMR spectrometer of the Bijvoet Center (Department for NMR Spectroscopy). In the case of WGA and UDA, only NOESY spectra in D2O at 300 K were recorded. Computational Calculations MD simulations for hevein, pseudohevein, and domB were carried out by using the CVFF (consistent valence force field).42,43 In the case of hevein and pseudohevein we also used the GROMOS (Groningen molecular simulations)44,45 force field for the molecular dynamics calculations at a simulation temperature of 300 K. The starting structures of the lectins were either taken from the Brookhaven Protein Data Bank or generated by using knowledge-based homology modeling technique. The generation of the H atoms, automatic assignment of partial charges for each atom of the molecule, was accomplished by using the INSIGHTII software. The center of mass of the lectin was placed in the center of a 30 3 30 3 30 Å large water box. A simple point charge model for the water was used, allowing the atoms of the water molecules to vibrate. Periodic boundary conditions were applied by using a double cutoff limit of 10 and 12 Å for the water box. Each molecular ensemble was submitted to a molecular dynamics simulation by using the CVFF force field at a temperature of 300 K, with an equilibration time of 20 ps, a production period of 100 ps, an integration step of 1 fs and a dielectric constant of 1. Each 250 steps of integration a conformation was stored. From the 400 individual conformations obtained within the production time the 10 275 CARBOHYDRATE RECOGNITION BY LECTINS TABLE VII. Surface Accessibilities [Å2] of CIDNP-Reactive Amino Acid Groups of 10 domB (WGA1) Structures Representing Relative Local Energy Minima According to MD Simulations* I II III IV V VI VII VIII IX X x 107.3 86.3 32.8 115.3 64.5 28.1 117.5 75.6 46.2 129.0 100.8 83.8 133.3 128.9 81.2 122.4 106.3 89.0 120.8 92.8 70.1 119.8 79.4 56.1 127.2 86.8 71.2 a. Start structure: ProMod based on hevein structure. Dot density: 1, sphere radius: 1.5 Å Tyr64 Tyr66 Tyr73 141.8 77.8 75.9 141.8 68.4 52.8 111.7 79.0 70.4 126.8 78.0 57.7 132.9 66.0 46.5 117.7 71.9 37.9 86.1 92.4 26.3 93.7 67.5 29.7 b. Start structure: ProMod based on all similar structures. Dot density: 1, sphere radius: 1.5 Å Tyr64 Tyr66 Tyr73 107.2 96.9 73.3 111.8 110.7 79.7 104.4 104.5 90.3 113.8 106.5 72.1 133.3 104.5 95.1 119.2 102.8 110.8 125.7 107.1 103.3 146.3 100.2 100.8 c. Start structure: X-ray data of domB from WGA1. Dot density: 1, sphere radius: 1.5 Å Tyr64 Tyr66 Tyr73 134.4 83.9 68.2 138.1 88.9 67.7 122.6 82.5 71.8 122.8 85.2 68.0 131.5 88.3 85.7 124.6 89.0 77.5 127.0 92.1 80.0 124.5 85.5 66.7 *MD simulations were carried out in the CVFF (consistence valence force field).42,43 with the lowest potential energy were automatically selected and processed for energy minimization of the complete ensemble by using the conjugate gradient method. Generation of possible conformations for domB was achieved by using different approaches: 1) singling out the domB from the X-ray structure of the complete lectin (PDB-entry 7WGA); and 2) knowledge-based homology modeling techniques by using the ProMod tools46 based on a) 1HEV as 3D template and b) 1WGC, 2WGC, 1WGT, and 1HEV as 3D template. The MD calculations were performed on an IBM-SP2 parallel machine by using four processors. The handling of the molecules and the analysis of the MD results were accomplished with various modules of the INSIGHTII software on a Silicon Graphics Indigo2 workstation. The surface accessibilities of the respective amino acid side chains were calculated by using the X-ray structures of WGA1 and WGA2 (Brookhaven Protein Data Bank) or MD structures of hevein and pseudohevein that have been ascertained independently by two-dimensional NMR measurements. The calculations were carried out with the help of the Connolly program in InsightII on a Silicon Graphics Personal Iris following an established method.47,48 The assessment of the routine ‘‘surface’’ based on the Connolly program49,50 pinpoints the accessible exterior part of the relevant portions of the molecule by smoothing the van der Waals surface with a test sphere that displays the average radius of the solvent water (1.5 Å) or the average radius of flavin (4 Å). Both average radii were calculated according to published data.51 The dot density was generally set to a value of 1. RESULTS AND DISCUSSION Hevein The small size of the protein allows unequivocal assignment of signals to individual protons and comparison of the NMR-derived structure with calculated surface accessibilities as quality control. The CIDNP signals of hevein indicate a high accessibility for Trp21 and for Trp23 in contrast to a low accessibility for Tyr30 (Fig. 1). This agrees well with calculations that assess the surface accessibilities of the respective residues in several hevein NMR conformations. For example, the surface display of the CIDNPreactive moieties was calculated for eight hevein energy-minimized conformations we derived from combined application of our own NMR data (unpublished results) and MD simulations by using the GROMOS force field (Table I). When this set was compared to seven published conformational variants of hevein,52 especially the binding region showed a high degree of conformational conservation. The calculated surface accessibilities of the CIDNPreactive amino acids in these resolvable spatial arrangements are compiled in Table II. The average radius of flavin (4 Å) was chosen in addition to that of water (1.5 Å) to reflect the fact that the CIDNP effect is generated by an interaction with the dye, as given in Tables I and II. In general, the calculated accessibilities of the tyrosine, tryptophan and histidine residues decrease with the assumed increase of the sphere radius. At both test sphere radii the same residues can be identified as having low, medium, or high accessiblity. For comparison, surface accessibility values for the different hevein conformations derived from MD calculations in the CVFF are also given in Table III. A significant difference between protein conformational aspects and surface accessibility was detected only for His35, which is positioned in a highly flexible part of the protein chain. By including the accessibility values derived from different energy minimum conformations into our 276 H.-C. SIEBERT ET AL. Fig. 6. Laser photo CIDNP difference spectra (aromatic part) of domB (WGA1) and domB (WGA1)–(GlcNAc)4 complexes. The small positive CIDNP signals originate from the d1, d2 protons of the tyrosine residues. a: 2 mmol domB (WGA1) 1 1 mmol (GlcNAc)4. b: 2 mmol domB (WGA1) 1 1.5 mmol (GlcNAc)4. c: CIDNP-difference spectrum of the aromatic part of the corresponding one-dimensional spectra, given above. 2 mmol domB (WGA1) 1 2 mmol (GlcNAc)4. tables, we take into account the inherent protein flexibility and the resulting fluctuation of the surface presentation of this side chain. It is evident that movement of side chains alters their surface exposure within certain limits. When the accessible area adopts a value above 80 Å2 (with a test sphere radius of 1.5 Å), the tyrosine, histidine, and tryptophan residues can legitimately be considered as surfaceexposed and are thereby able to generate a CIDNP signal of high intensity. When chitobiose is added to hevein, CIDNP signals are altered. The intensities of the signals of the protons, whose reactivities to the photoactivated dye are affected by ligand binding, decrease (Fig. 1). As can be seen in Figures 1 and 2, the overlapping h2-proton signals of Trp21 and Trp23 are broadened but their intensity is not weakened. In Fig. 2c one can see that the intensity of the h2-proton signal is even enhanced. These results indicate that the h2-protons are not strongly shielded by the ligand, which means that distinct parts of the CARBOHYDRATE RECOGNITION BY LECTINS 277 Fig. 7. Energy minimum conformation of domB of WGA1, emphasizing the surface exposition of Tyr64 and Tyr66 in the upper part of the figure. Trp residues are covered differently. Such an observation can be helpful in case of computer-assisted docking studies because a defined reorientation of the two Trp residues in case of ligand binding has been detected. Our results thus reveal a carbohydrate-dependent effect on the CIDNP spectrum, which in the absence of a major conformational change by ligand binding can be attributed to ligandshielding of certain amino acid side chains. Involvement of Ser19, Trp21, Trp23, and Tyr30 in carbohydrate binding has been proven independently by two-dimensional NMR experiments53 (unpublished results). The availability of NMR-derived coordinates of the hevein–chitobiose complex, kindly provided in detail by Dr. Jimenez-Barbero, allows comparison of surface accessibility values of this complex in solution to the corresponding values of hevein in the absence of the ligand, as given in Table IV. These results corroborate the evidence from the CIDNP experiments that the ligand reduces the extent of interaction between these residues and the dye. CIDNP experiments at two different pH values were carried out in order to test whether the surface accessibility is affected by this parameter. An apparent pH dependence of the hevein conformation was indicated by the change in intensity of the signals of Trp21, Trp23, Tyr30, and His35 (Figs. 1, 2). En- hanced intensity at increasing pH values can be attributed to an increased accessibility of Tyr30 due to conformational changes at elevated pH values. This interpretation is in agreement with results of a recent study, in which the pH dependence of the conformation in a distinct part of hevein, that is, the part between amino acid residues 28 and 35, has been described.52 Pseudohevein For pseudohevein, where the highly exposed and flexible Trp21 is a tyrosine moiety, we observed a strong Tyr signal in the CIDNP spectrum (Fig. 3a). The CIDNP signal of the conserved Trp23 is maintained. The different degrees of accessibility as illustrated by the CIDNP data could be confirmed by molecular dynamics calculations, compiled in Tables V and VIa,b. During the simulation period, 10 intermediate coordinate frames were collected and energetically minimized. The initial structure was derived from an energy-minimized coordinate frame of hevein in which the six variant amino acids were replaced by the appropriate residues of pseudohevein. The resulting coordinate frame was energetically minimized. Possible differences between the conformations of hevein and pseudohevein were analyzed by MD simulations. The MD calculations of 278 H.-C. SIEBERT ET AL. Fig. 8. Laser photo CIDNP difference spectra (aromatic part) of UDA and UDA-(GlcNAc)4 complexes. a: Ligand-free UDA. b: 1 mmol UDA 1 1 mmol (GlcNAc)4. pseudohevein indicate that Tyr21 has a significantly higher accessibility than Tyr30, independent of the applied force field, either GROMOS (Table V) or CVFF (Table VIa,b). However, comparisons between the accessibility values given in Tables V, VIa,b show remarkable differences. Taking the strong CIDNP signals from tyrosine into account, the values of Table VIb obtained with the ProMod procedure yield the closest agreement to the experimental data (Fig. 3a,b). A low-energy minimum conformation of a complete pseudohevein molecule, depicted in Figure 4, illustrates the high degree of surface exposure of Tyr21 and Trp23. The surface exposure of Tyr30 is increased relative to hevein (Table VIb). The CIDNP spectrum of the pseudohevein–(GlcNAc)4 complex (Fig. 3b) suggests an involvement of residues 21, 23, and 30 in carbohydrate binding, which is consistent with the results obtained for hevein. To illustrate the quantitative relationship between signal intensity and ligand concentration, the degree of suppression of the nonoverlapped signal of proton Trp23e3 in dependence of the ligand concentration is shown in Figure 3c. Moreover, the pH dependence of conformational features of pseudohevein is similar to that of hevein (not shown). It is remarkable that pseudohevein and hevein do not lose their binding abilities, not even at pH 1, despite the presence of conformational changes. Having monitored the CIDNP behav- ior of these two small lectins, we proceeded to use a single domain of the more complex agglutinin WGA, obtained from genetic engineering, in the next series of measurements, extending the experimental basis to more complex GlcNAc-binding lectins of this family. Domain B of Wheat Germ Agglutinin To answer the question whether the so far documented ligand-dependent signal suppression of CIDNP-reactive amino acids will also be observed in a single heveinlike domain of WGA1, we investigated the recombinant domain B (domB). Interpretation of the CIDNP signals of the complete WGA molecule can be more easily analyzed, when spectra of the single domain are available. The CIDNP spectra of ligand-free domB display overlapping signals for Tyr66e1,2 and Tyr73e1,2 overlap (Fig. 5). Tyr64e1,2 generates the most pronounced CIDNP signal which points to a very high accessibility (Fig. 5). This observation deserved special attention in the molecular modeling studies of domB. One initial structure was derived from a low-energy minimum coordinate frame of hevein. The second start structure was obtained by comparison with all other similar structures by using the ProMod procedure.46 The third start structure was taken from the X-ray data of WGA1 in the Brookhaven protein data bank. CARBOHYDRATE RECOGNITION BY LECTINS Fig. 9. Laser photo CIDNP difference spectra (aromatic part) of WGA2 and WGA2-(GlcNAc)4 complexes. a: Ligand-free WGA2. b: 0.5 mmol WGA2 1 0.5 mmol (GlcNAc)4. c: 0.5 mmol WGA2 1 0.75 mmol (GlcNAc)4. d: 0.5 mmol WGA2 1 1 mmol (GlcNAc)4. e: Signal intensity of Tyr in dependence of the ligand concentration. 279 280 H.-C. SIEBERT ET AL. Urtica dioica Agglutinin Figure 9. (Continued.) All MD simulations demonstrated that Tyr64 has the highest accessibility (Table VIIa,b,c). As observed for hevein and pseudohevein, the CIDNP signals of domB can be suppressed by addition of N-acetylglucosamine oligomers. When the concentration of domB in CIDNP studies was increased (Fig. 6), all three Tyre1,2 signals were separated at a 2:1 molar ratio between domB and (GlcNAc).4 The three Tyr-dependent CIDNP signals behaved differently as a function of the ligand (Fig. 5). The CIDNP signal of Tyr64e1,2 was most strongly affected when (GlcNAc)4 was added to domB in a molar ratio of 1:2. By contrast, the CIDNP signals of Tyr66e1,2 and Tyr73e1,2 decreased after further enhancement of the ligand concentration and finally were suppressed more strongly than the Tyr64e1,2 signal. This result illustrates that it is possible to detect different degrees of responsiveness of CIDNP-sensitive residues. The ligand-dependent CIDNP reactivity of all three Tyr residues underscores their role in stabilizing protein–carbohydrate complexes. The relative orientations of three Tyr residues are modeled in Figure 7. Based on these results, we conclude that all three structure-generating procedures lead to comparable conformations of the binding region that are in agreement with the CIDNP data, although distinct differences in surface accessibility of Tyr73, for example, occur. These findings are of special interest because deviations between the X-ray structure and NMR structures of hevein have been reported.52,54 Having so far evaluated the properties of proteins with one GlcNAc-binding site, the next two paragraphs will report data on agglutinins with two or four domains. UDA consists of two hevein domains. Alignment of UDA with the other heveinlike lectins shows that Trp21, Trp23, and Tyr30 as well as His67, Trp69, and Tyr76 are the corresponding aromatic amino acids in the binding pocket. Trp16, Trp40, His47, and Tyr84 are other CIDNP-reactive amino acid residues of UDA, which are located in exposed surface regions other than the GlcNAc-binding sites.26 CIDNP spectra of hevein–, pseudohevein– or domB–carbohydrate complexes show a reduced signal intensity in relation to the CIDNP spectra of ligand-free protein (Figs. 1–3, 5, 6). In contrast to this, the CIDNP spectrum of UDA with its two overlapping Tyr signals is not strongly affected after ligand addition (Fig. 8). This points to a lesser extent of shielding of the binding site residues of UDA by (GlcNAc)– oligomers in comparison to that observed for the hevein, pseudohevein and domB binding sites, which show reduced signals. Assignment of the signals is in agreement with published data.55 The occurrence of the Trp21e3 signal in the CIDNP spectra of UDA– (GlcNAc)–oligomer complexes raises the question whether GlcNAc binding is accompanied by a conformational change of and/or around Trp21 (Fig. 8). A similar observation has been made based on a biochemical experiment in the case of an insect lectin, whose positive surface charge and susceptibility to trypsin are altered by lactose binding.56 The presence of a ligand-induced conformational change allows consideration of various models. In one of these models a significant conformational change of the carbohydrate receptor has previously been predicted.57,58 In view of the remarkable homology to hevein, it is notable that the resonances of Trp21 and Trp23 of UDA are broadened in the presence of the ligand.55 This homology to hevein is obviously indicated also by the similarities of the CIDNP spectra of hevein (Fig. 1) and that of UDA (Fig. 8). Wheat Germ Agglutinin The WGA monomer consists of four hevein domains (A, B, C, D), each with an architecturally unique binding site characterized by aromatic amino acid residues. These residues are tyrosines in the binding sites of domains A and B (Tyr21, -23, -30; and Tyr64, -66, and -70, respectively). In WGA2 Tyr66 is replaced by His. Domains C and D have a somewhat different character where the corresponding residues are Trp107, Phe109, and Phe116 as well as Trp150 and Tyr159, respectively. A conserved Ser residue, which interacts with the carbonyl group of bound GlcNAc, is present in all four binding sites. A subsidiary binding region located on an adjacent domain of the second monomer further contributes to ligand stabilization in all binding sites except that of domain A. Thus, as an example of a more complex structure, WGA can be used to test the conclusions 281 CARBOHYDRATE RECOGNITION BY LECTINS drawn from the simple one- or two-domain lectins. As illustrated in Figure 9a, several surface-exposed aromatic residues (His, Trp, Tyr) of WGA2 produce signals in the difference CIDNP spectrum. Upon ligand binding a decrease of the intensity of the Tyr signal is observed as a function of the oligosaccharide chain length (N,N8-diacetylchitobiose, N,N8,N9triacetylchitotriose or N,N8,N9,N--tetraacetylchitotetraose) (not shown) and ligand concentration (Fig. 9a–d). The gradual decrease of the peak areas as a function of increases in saccharide concentration is shown in Fig. 9b–d. When the measured intensities are plotted against the ligand concentration, the pronounced shielding of the ligand is readily apparent (Fig. 9e). Remarkably, it is evocative of the correlation of these two parameters determined in the case of pseudohevein (Fig. 3c). The Trp/His signals are also strongly reduced after stepwise addition of N,N8,N9,N--tetraacetylchitotetraose to WGA2 as demonstrated in Fig. 9b–d. In order to identify which aromatic amino acid side chains can contribute most to the intensity of the CIDNP signals, we have calculated the surface accessibilities of the tyrosine, tryptophan and histidine residues from the available crystal structures (Brookhaven protein data bank), as shown in Table VIII. It is apparent from these values that Tyr21 and Tyr64 possess the highest accessibility and presumably contribute most strongly to the tyrosine CIDNP signal intensity in the difference spectrum. The surface accessible Tyr residues generate only one Tyr CIDNP signal at 6.8 ppm which decreases at a slower rate with increasing ligand-concentration (Fig. 9a–d) than the CIDNP signals of the Trp/His residues do. This is in accordance with the fact that four tyrosine moieties are present in the two weak binding sites (3 in site A and one in site D) and that these sites may not be fully occupied at low ligand concentration. As mentioned above, histidine is present only in the B binding site of WGA2, and its accessibility is rather low (Table VIII). A highly accessible Trp107 appeared to be present in the binding site in the C-domain of WGA. Although GlcNAc binding at the C-domain site was not observed in the crystal due to molecular packing interactions, binding of NeuNAc was obtained in another crystal form in which it participates in a crosslinking binding mode. The evidence from our CIDNP spectra suggests that GlcNAc–oligomers occupy the C-domain site due to the high accessibility of Trp107 and the strong ligand-dependent responsiveness of the Trp/His signal (Fig. 9). The CIDNP technique thus provides evidence for occupancy of all four sites at high ligand concentration. However, the spectrum is complicated by the fact that aromatic amino acids that are not located in the binding sites also contribute to the CIDNP signals (His59, Trp41, Tyr145). Thus, the resonances of the aromatic amino TABLE VIII. Surface Accessibilities [Å2] of CIDNP-Reactive Amino Acid Groups I II a. WGA1 (monomer I, monomer II). Dot density: 1, sphere radius: 1.5 Å Tyr21 Tyr23 Tyr30 Trp41 Tyr64 Tyr66 Tyr73 Trp107 Tyr145 Trp150 Tyr159 96.6 94.0 59.8 136.0 111.4 69.5 63.8 125.7 55.6 56.7* 62.0 103.7 96.3 43.3 132.3 102.1 82.4 52.2 121.9 57.9 142.2 73.6 b. WGA2 (monomer I, monomer II) according to the X-ray structures taken from the Protein Data Bank. Dot density: 1, sphere radius: 1.5 Å Tyr21 Tyr23 Tyr30 Trp41 His59 Tyr64 His66 Tyr73 Trp107 Tyr145 Trp150 Tyr159 91.2 98.0 46.7 131.4 44.3 97.9 50.1 55.4 142.0 50.2 161.0 58.1 96.2 87.2 61.5 130.5 31.6 101.5 57.9 62.9 136.3 40.8 55.5* 69.8 *The complete structural coordinates of this residue are not given in the Protein Data Bank. acid residues are not quenched in the presence of the specific ligand (Fig. 9). With regard to the binding of NeuNAc oligosaccharide or 2-deoxy-2,3-didehydro-N-acetylneuraminic acid (Neu5Ac2en) (data not shown) to WGA, our method fails to detect any changes in the CIDNP spectrum relative to the ligand-free lectin. This result suggests either that NeuNAc covers the binding pocket differently or that the method is not suitable to analyze weak interactions of a ligand to the binding site. Commercially available isolectin mixtures of WGA exhibit the same changes in the CIDNP signals, as observed in either the WGA1 or WGA2 spectrum, when a ligand is added (not shown). This observation suggests that both isoforms of WGA behave similarly upon ligand binding. 282 H.-C. SIEBERT ET AL. CONCLUSIONS GlcNAc-specific lectins of increasing structural complexity have been analyzed in CIDNP experiments monitoring the surface exposure of histidine, tyrosine and tryptophan as sensors for carbohydrateprotein interaction. It could be demonstrated that carbohydrate binding reduces the intensity of distinct signals due to ligand shielding of several aromatic amino acids in the binding regions. In the case of UDA the surface accessibility of a Trp residue increases as a result of addition of GlcNAc–oligomers, suggesting a ligand-induced change of surface presentation of that residue. The single domain lectins hevein and pseudohevein and recombinant domB of WGA1 show a similar pattern of responses, corroborating the notion that aromatic amino acids are present in all homologous binding sites. In addition to ligand-dependent differences in signal intensity variations of the pH value can also lead to conformational changes, which, however, appear not to affect ligand binding. Since only three types of aromatic amino acids are CIDNP-sensitive, the probability of signal overlap in a CIDNP difference spectrum is lower than in the conventional one-dimensional 1H-NMR spectrum. Furthermore, as shown in Fig. 6c, an aromatic amino acid signal in an one-dimensional 1H-NMR spectrum can be broadened and/or have reduced intensity in the presence of a saccharide-ligand. The CIDNP approach allows correlation of signal intensity and signal responsiveness as a function of surface exposure and ligand shielding of a specific residue. These measurable parameters can therefore be helpful in optimizing molecular models for computer-aided docking studies. As also revealed in this study, it is possible to comparatively assess the validity of computer-assisted structural calculations by different modeling techniques. The significant structural differences between the pentraxins serum amyloid P component and C-reactive protein despite their remarkably high degree of sequence homology call attention to the necessity to take any reliable conformational parameter into account for meaningful modeling studies to avoid being led astray.59 In sum, the CIDNP method provides a valuable tool to study protein–carbohydrate interaction in addition to multidimensional NMR-techniques, Xray-analysis and computer-assisted molecular modeling.16,60–65 ACKNOWLEDGMENTS We are indebted to Dr. N. H. Andersen and Dr. B. Cao for information on their hevein NMR data, Dr. J. Jimenez-Barbero for information on his hevein– ligand complex NMR data, Dr. L. Kroon-Batenburg for advice to perform calculations of surface accessibility, and P. Edebrink and J. O. Koopmann for technical assistance. This work was supported by the Human Capital and Mobility Program of the European Community, and the Dr. M. Scheel-Stiftung für Krebsforschung. REFERENCES 1. Sharon, N., Lis, H. Lectins as cell recognition molecules. Science 246:227–234, 1989. 2. Gabius, H.-J. Detection and functions of mammalian lectins: With emphasis on membrane lectins. Biochim. Biophys. Acta 1071:1–18, 1991. 3. Gabius, H.-J., Gabius, S. (eds.) ‘‘Lectins and Glycobiology.’’ New York: Springer-Verlag, 1993. 4. Gabius, H.-J., Gabius, S. (eds.) ‘‘Glycosciences: Status and Perspectives.’’ Weinheim: Chapman & Hall, 1997. 5. Kaptein, R., Dijkstra, K., Nicolay, K. Laser photo-CIDNP as a surface probe for proteins in solution. Nature 274:293– 294, 1978. 6. Kaptein, R. Photo CIDNP studies of proteins. In: ‘‘Biological Magnetic Resonance,’’ Vol. 4. Berliner, L. J. (ed.). New York: Plenum, 1982:145–191. 7. Hore, P.J., Broadhurst, R.W. Photo-CIDNP of biopolymers. Prog. NMR Spectrosc. 25:345–402,1993. 8. Buck, F., Rüterjans, H., Kaptein, R., Beyreuter, K. Photochemically induced dynamic nuclear polarization investigation of complex formation of the NH2-terminal DNAbinding domain of lac repressor with poly [d(AT)]. Proc. Natl. Acad. Sci. U.S.A. 77:1545–1548, 1980. 9. Stob, S., Scheek, R. M., Boelens, R., Kaptein, R. PhotoCIDNP study of the interaction between lac repressor headpiece and lac operator DNA. FEBS Lett. 239:99–104, 1988. 10. Broadhurst, R.W., Dobson, C.M., Hore, P.J., Radford, S.E., Rees, M.L. A photochemically induced dynamic nuclear polarization study of denaturated states of lysozyme. Biochemistry 30:405–412, 1991. 11. Hård, K., Kamerling, J.P., Vliegenthart, J.F.G. Application of laser photo-CIDNP for an intact glycoprotein in solution. Carbohydr. Res. 236:321–326, 1992. 12. Siebert, H.-C., André, S., Reuter, G., Gabius, H.-J., Kaptein, R., Vliegenthart, J.F.G. Effect of enzymatic desialylation of human serum amyloid P component on surface exposure of laser photo CIDNP (chemically induced dynamic nuclear polarization)-reactive histidine, tryptophan and tyrosine residues. FEBS Lett. 371:13–16, 1995. 13. Wright, C.S. Structural comparison of the two distinct sugar-binding sites in wheat germ agglutinin isolectin II. J. Mol. Biol. 178:91–104, 1984. 14. Wright, C.S. Crystal structure of a wheat germ agglutinin/ glycophorin-sialoglycopeptide receptor complex. J. Biol. Chem. 267:14345–14352, 1992. 15. Quiocho, F. Protein-carbohydrate interactions: Basic molecular features. Pure Appl. Chem. 61:1293–1306, 1989. 16. Imberty, A., Bourne, Y., Cambillau, C., Rouge, P., Perez, S. Oligosaccharide conformation in protein/carbohydrate complexes. Adv. Biophys. Chem. 3:71–117, 1993. 17. Gabius, H.-J. Animal lectins. Eur. J. Biochem., 243:543– 576, 1997. 18. Martins, J. C., Maes, D., Loris, R., Pepermans, H.A.M., Wyns, L., Willem, R., Verheyden, P. 1H NMR study of the solution structure of Ac-AMP2, a sugar binding antimicrobial protein isolated from Amaranthus caudatus. J. Mol. Biol. 258:322–333, 1996. 19. Soedjanaatmadja, U.K.S., Hofsteenge, J., JeronimusStratingh, C.M., Bruins, A.P. Beintema, J.J. Demonstration by mass spectrometry that pseudo-hevein and hevein have ragged C-terminal sequences. Biochim. Biophys. Acta 1209:144–148, 1994. 20. Soedjanaatmadja, U.K.S., Subroto, T., Beintema, J.J. The effluent of natural rubber factories is enriched in the antifungal protein hevein. Bioresource Technol. 53:39–41, 1995. 21. Chrispeels, M.J., Raikhel, N.V. Lectins, lectin genes, and their role in plant defence. Plant Cell 3:1–9, 1991. 22. Gidrol, X., Chrestin, H., Tan, H.-L., Kush, A. Hevein, a CARBOHYDRATE RECOGNITION BY LECTINS 23. 24. 25. 26. 27. 28. 29. 30. 31. 32. 33. 34. 35. 36. 37. 38. 39. 40. 41. 42. 43. lectin-like protein from Hevea brasiliensis (rubber tree), is involved in coagulation of latex. J. Biol. Chem. 269:9278– 9283, 1994. Peumans, W.J., van Damme, E.J.M. Lectins as plant defence proteins. Plant Physiol. 109:347–352, 1995. Rüdiger, H. Structure and function of plant lectins. In: ‘‘Glycosciences: Status and Perspectives.’’ Gabius, H.-J., Gabius, S. (eds.) Weinheim: Chapman & Hall, 1997:415– 438. Shibuya, N., Goldstein, I.J., Schafer, J.A., Peumans, W.J., Broeckaert, W.F. Carbohydrate binding properties of the stinging nettle (Urtica dioica) rhizome lectin. Arch. Biochem. Biophys. 249:215–224, 1986. Beintema, J.J., Peumans, W.J. The primary structure of stinging nettle (Urtica dioica) agglutinin. FEBS Lett. 299:131–134, 1992. Galelli, A., Truffa-Bachi, P. Urtica dioica agglutinin: A superantigenic lectin from stinging nettle rhizome. J. Immunol. 151:1821–1831, 1993. Gabius, H.-J. Lectinology meets mythology: Oncological future for the mistletoe lectin? Trends Glycosci. Glycotechnol. 6:229–238, 1994. Gabius, H.-J., Gabius, S., Joshi, S.S., Koch, B., Schroeder, M., Manzke, W. M., Westerhausen, M. From ill-defined extracts to the immunomodulatory lectin: Will there be a reason for oncological application of mistletoe? Planta Med. 60:2–7, 1994. Gabius, H.-J., Kayser, K., Gabius, S. Protein-ZuckerErkennung. Grundlagen und medizinische Anwendungen am Beispiel der Tumorlektinologie. Naturwissenschaften 82:533–543, 1995. Wright, H.T., Brooks, D.M., Wright, C.S. Evolution of the multidomain protein wheat germ agglutinin. J. Mol. Evol. 21:133–138, 1985. Wright, C.S., Raikhel, N. Sequence variability in three wheat germ agglutinin isolectins: Products of multiple genes in polyploid wheat. J. Mol. Evol. 28:327–336, 1989. Wright, H.T., Sandrasegaram, G., Wright, C.S. Evolution of a family of N-acetylglucosamine-binding proteins containing the disulfide-rich domain of wheat germ agglutinin. J. Mol. Evol. 33:283–294, 1991. Wright, C.S. Refinement of the crystal structure of wheat germ agglutinin isolectin 2 at 1.8 Å resolution. J. Mol. Biol. 194:501–529, 1987. Wright, C.S. Comparison of the refined crystal structure of two wheat germ agglutinin isolectins. J. Mol. Biol. 209:475– 487, 1989. Wright, C.S. 2.2 Å resolution structure analysis of two refined N-acetylneuraminyllactose-wheat germ agglutinin isolectin complexes. J. Mol. Biol. 215:635–651, 1990. Wright, C.S., Jaeger, J. Crystallographic refinement and structure analysis of the complex of wheat germ agglutinin with a bivalent sialoglycopeptide from glycophorin A. J. Mol. Biol. 232:620–638, 1993. Rice, A.C. ‘‘Cloning, Expression, Purification and Characterization of Domain B of Wheat Germ Agglutinin.’’ Doctoral thesis, Virginia Commonwealth University, Richmond, VA, USA, 1994. Kayser, K., Bovin, N.V., Zemlyanukhina, T.V., DonaldoJacinto, S., Koopmann, J., Gabius, H.-J. Cell typedependent alterations of binding of synthetic blood group antigen-related oligosaccharides in lung cancer. Glycoconjugate J. 11:339–344, 1994. Wright, C.S. Histidine determination in wheat germ agglutinin isolectin by X-ray diffraction analysis. J. Mol. Biol. 145:453–461, 1981. Zheng, J., Schodel, F., Peterson, D.L. The structure of hepadnaviral core antigens. J. Biol. Chem. 267:9422–9429, 1992. Hagler, A.T., Huler, E., Lifson, S. Energy functions for peptides and proteins: Deviation of a consistent force field including the hydrogen bond for amide crystals. J. Am. Chem. Soc. 96:5316–5327, 1974. Hagler, A.T., Lifson, S., Dauber, P. Consistent force field studies of intermolecular forces in hydrogen-bonded crystals. 2. A benchmark for the objective comparison of 44. 45. 46. 47. 48. 49. 50. 51. 52. 53. 54. 55. 56. 57. 58. 59. 60. 61. 62. 283 alternative force fields. J. Am. Chem. Soc. 101:5122–5130, 1979. van Gunsteren, W.F. ‘‘GROMOS, Groningen Molecular Simulation Package.’’ University of Groningen, The Netherlands, 1987. van Gunsteren, W.F. The role of computer simulation techniques in protein engineering. Protein Eng. 2:5–13, 1988. Peitsch, M.C. ProMod and Swiss-Model: Internet-based tools for automated comparative protein modelling. Biochem. Soc. Transact. 24:274–279, 1996. Lee, B., Richards, F.M. The interpretation of protein structures: Estimation of static accessibility. J. Mol. Biol. 55:379– 400, 1971. Kellenbach, E., Härd, T., Boelens, R., Dahlman, K., Carlstedt-Duke, J., Gustafsson, J-Å., van der Marel, G. A., van Boom, J.H., Maler, B., Yamamoto, K.R., Kaptein, R. PhotoCIDNP study of the interaction between the glucocorticoid receptor DNA-binding domain and glucocorticoid response elements. J. Biomol. NMR 1:105–110, 1991. Connolly, M.L. Analytical molecular surface calculation. J. Appl. Crystallogr. 16:548–558, 1983. Connolly, M.L. Solvent-accessible surfaces of proteins and nucleic acids. Science 221:709–713, 1983. Edward, J.T. Molecular volumes and the Stokes-Einstein equation. J. Chem. Educ. 47:261–270, 1970. Andersen, N.H., Cao, B., Rodriguez-Romero, A., Arreguin, B. Hevein: NMR assignment and assessment of solutionstate folding for the agglutinin-toxin motif. Biochemistry 32:1407–1422, 1993. Asensio, J.L., Canada, F.J., Bruix, M., Rodriguez-Romero, A., Jimenez-Barbero, J. The interaction of hevein with N-acetylglucosamine-containing oligosaccharides: Solution structure of hevein complexed with chitobiose. Eur. J. Biochem. 230:621–633, 1994. Rodriguez-Romero, A., Ravichandran, K.G., SorianaGarcia, M. Crystal structure of hevein at 2.8 Å resolution. FEBS Lett. 291:307–309, 1995. Hom, K., Gochin, M., Peumans, W.J., Shine, N. Ligandinduced perturbations in Urtica dioica agglutinin. FEBS Lett. 361:157–161, 1995. Komano, H., Kurama, T., Nagasawa, Y., Natori, S. Evidence for an increase in positive surface charge and an increase in susceptibility to trypsin of Sarcophaga lectin (from flesh fly, Sarcophaga peregrina) on its interaction with galactose, a hapten sugar of the lectin. Biochem. J. 284:227–230, 1992. Efting, M.R., Anusiem, A.C., Biltonen, R.L. Enthalpyentropy compensation and heat capacity changes for protein-ligand interactions: General thermodynamic models and data for the binding of nucleotides to ribonuclease A. Biochemistry 22:3884–3896, 1983. Perez, S., Imberty, A., Carver, J. Molecular modeling: An essential component in the structure determination of oligosaccharides and polysaccharides. Adv. Comput. Biol. 1:147–202, 1994. Shrive, A.K., Cheetham, G.M.T., Holden, D., Myles, D.A.A., Turnell, W.G., Volanakis, J.E., Pepys, M.B., Bloomer, A.C., Greenhough, T.J. Three dimensional structure of human C-reactive protein. Nature Struct. Biol. 3:346–354, 1996. Siebert, H.-C., Kaptein, R., Vliegenthart J.F.G. Study of oligosaccharide-lectin interaction by various nuclear magnetic resonance (NMR) techniques and computational methods. In: ’’Lectins and Glycobiology.’’ Gabius, H.-J., Gabius, S. (eds.). New York: Springer-Verlag, 1993:105–116. Siebert, H.-C., Gilleron, M., Kaltner, H., von der Lieth, C.-W., Kozàr, T., Bovin, N.V., Korchagina, E.Y., Vliegenthart, J.F. G., Gabius H.-J. NMR-based, molecular dynamics and random walk molecular mechanics-supported study of conformational aspects of a carbohydrate ligand (Galb1,2Galb1-R) for an animal galectin in the free and in the bound state. Biochem. Biophys. Res. Commun. 219:205– 212, 1996. Siebert, H.-C., von der Lieth, C.-W., Gilleron, M., Reuter, G., Wittmann, J., Vliegenthart, J.F.G., Gabius, H.-J. Carbohydrate-protein interaction. In: ‘‘Glycosciences: Status and 284 H.-C. SIEBERT ET AL. Perspectives.’’ Gabius, H.-J., Gabius, S. (eds.). Weinheim: Chapman & Hall. 1997:291–310. 63. Siebert, H.-C., von der Lieth, C.-W., Dong, X., Reuter, G., Schauer, R., Gabius, H.-J., Vliegenthart, J.F.G. Molecular dynamics-derived conformation and intramolecular interaction analysis of the N-acetyl-9-O-acetylneuraminic acidcontaining ganglioside GD1a and NMR-based analysis of its binding to a human polyclonal immunoglobulin G fraction with selectivity for O-acetylated sialic acids. Glycobiology 6:561–572, 1996. 64. van Halbeek, H. NMR development in structural studies of carbohydrates and their complexes. Curr. Opin. Struct. Biol. 4:697–709, 1994. 65. von der Lieth, C.-W., Kozár, T., Hull, W.E. A (critical) survey of modeling protocols used to explore the conformational space of oligosaccharides. J. Mol. Struct. (Theochem), in press, 1997. 66. Brooks, R.B., Bruccoleri, R.E., Olafson, B.D., States, D.J., Swaminathan, S., Karplus, M. CHARMM: A program for macromolecular energy minimization and dynamic calculations. J. Comput. Chem.4:187–217, 1983.