Structures of the Erythrina corallodendron lectin and of its complexes with mono- and disaccharides1

J. Mol. Biol. (1998) 277, 917±932

Structures of the Erythrina corallodendron Lectin and of its Complexes with Mono- and Disaccharides Sharona Elgavish and Boaz Shaanan* Department of Biological Chemistry and The Wolfson Centre for Applied Structural Biology, The Institute of Life Sciences, The Hebrew University of Jerusalem Givat Ram, Jerusalem 91904 Israel

The structures of the Erythrina corallodendron lectin (EcorL) and of its complexes with galactose, N-acetylgalactosamine, lactose and N-acetyllacÊ . The ®nal tosamine were determined at a resolution of 1.9 to 1.95 A R-values of the ®ve models are in the range 0.169 to 0.181. The unusual, non-canonical, dimer interface of EcorL is made of b-strands from the two monomers, which face one another in a ``hand-shake'' mode. The galactose molecule in the primary binding site is bound in an identical way in all four complexes. Features of the electrostatic potential of the galactose molecule match those of the potential in the combining site, thus probably pointing to the contribution of the electrostatic energy to determining the orientation of the ligand. No conformational change occurs in the protein upon binding the ligand. Subtle variations in the binding mode of the second monosaccharide (glucose in the complex with lactose and N-acetylglucosamine in the complex with N-acetyllactosamine) were observed. The mobility of Gln219 is lower in the complexes with the disaccharides than in the complexes with the monosaccharides, indicating further recruitment of this residue to ligand binding through more extensive hydrogen bonding in the former complexes. Water molecules that have been located in the combining sites of the ®ve structures undergo rearrangement in response to binding of the different ligands. The new structural information is in qualitative agreement with thermodynamic data on the binding to EcorL. # 1998 Academic Press Limited

*Corresponding author

Keywords: lectins; protein-carbohydrate interactions; carbohydrate speci®city

Introduction Speci®c recognition of carbohydrates by lectins underlies a variety of cellular processes such as cell-cell and host-pathogen interactions, lymphocyte homing, targeting of proteins into specialized cellular compartments and tissue development Abbreviations used: EcorL, Erythrina corallodendron lectin; Gal, galactose; Glc, glucose; Man, mannose; GlcNac, N-acetylglucosamine; GalNac, N-acetylgalactosamine; GS4, fourth lectin from Griffonia simplicifolia; PNA, peanut agglutinin; SBA, soybean agglutinin; PL, pea lectin; LOL I, Lathyrus ochrus isolectin I; ECPRO, free form of EcorL; ECGAL, ECNGAL, ECLAC and ECNAL, complexes of EcorL with Gal, N-acetylgalactosamine, lactose and N-acetyllactosamine, respectively; MPD, 2,4methylpentane-diol; NMR, nuclear magnetic resonance; r.m.s., root-mean-square; e.s.d., estimated standard deviation; PDB, Brookhaven Protein Data Base. 0022±2836/98/140917±16 $25.00/0/mb981664

(Sharon & Lis, 1990). In recent years, crystallographic studies have provided invaluable information on the structural aspects of lectincarbohydrate recognition and on the origins of their carbohydrate speci®city (for reviews, see Rini (1995) and Weis & Drickamer (1996)). The lectin isolated from seeds of the Erythrina corallodendron plant (EcorL) belongs to the family of legume lectins, which is one of the most extensively studied group of lectins. EcorL was the ®rst galactose-speci®c legume lectin for which the three-dimensional structure has been determined crystallographically (Shaanan et al., 1991). The principles of galactose (Gal) recognition by the legume lectins were established in that study, as has been con®rmed in subsequent crystallographic studies on the lectins from Griffonia simplicifolia (GS4; Delbaere et al., 1993), peanut (PNA; Banerjee et al., 1996) and soybean (SBA; Dessen et al., 1995). # 1998 Academic Press Limited

918 A prominent feature of lectins in general is their relatively low af®nity (in the millimolar range) to monosaccharides but increased af®nity to oligosaccharides (Rini, 1995; Weis & Drickamer, 1996). In the legume lectin family, monosaccharides bind in what is known as the primary binding site, typically a shallow depression on the surface of the protein. Additional carbohydrate units of the oligosaccharide chain interact with residues in the vicinity of the primary site, in the extended site or sub-site, and thus contribute to the increased overall speci®city (Rini, 1995; Weis & Drickamer, 1996). The interactions of the carbohydrate with residues in the combining site are either direct or mediated by water molecules (Weis & Drickamer, 1996). Rearrangement of water molecules in the combining site, following ligand binding, has been observed in several legume lectins, e.g. the pea lectin (PL; Rini et al., 1993) and GS4 (Delbaere et al., 1993) and identi®ed as an important contributor to the overall energetic balance of the binding process (Chervenak & Toone, 1994; Toone, 1994). Small changes in the structure of the combining site have been observed in the case of lectins speci®c to glucose/mannose (Glc/Man, respectively) such as PL (Rini et al., 1993), lentil lectin (Loris et al., 1994) and the Lathyrus ochrus lectin I (LOL I; Bourne et al., 1990b), but not in the Gal-binding lectin GS4 (Delbaere et al., 1993). Thus, a variety of ®nely tuned effects seem to govern ligand binding to legume lectins. A comprehensive understanding of the factors affecting the binding process can be gained by correlating information derived from high-resolution structural studies with thermodynamic data (Chervenak & Toone, 1994; Toone, 1994). We present the high-resolution structures of EcorL in complex with the monosaccharides galactose (ECGAL) and N-acetylgalactosamine (ECNGAL), the disaccharides N-acetyllactosamine (ECNAL) and lactose (ECLAC), as well as a structure of the free form of the lectin (ECPRO). A preliminary account of the ECLAC structure has been presented elsewhere (Shaanan et al., 1991). The fully re®ned structure of ECLAC is discussed here in the context of the other complexes and the free form of EcorL. In analysing the structures, we focus on the interactions stabilizing the carbohydrates in the primary and the extended binding sites, on the rearrangement of the water molecules and on the similarity between the interactions of the disaccharides lactose and N-acetyllactosamine with other lectins. While there is no conformational change between the unliganded and liganded forms of EcorL, we point out some changes of mobility in the combining site. We attempt to correlate the novel structural information on the unliganded and liganded states of EcorL with the available thermodynamic data on binding to EcorL (Surolia et al., 1996).

Structures of the Erythrina corallodendron Lectin

Results and Discussion Accuracy of structures Four of the ®ve structures (ECPRO, ECGAL, Ê ECLAC and ECNAL) were determined at 1.95 A Ê resolution. resolution and one (ECNGAL) at 1.9 A The agreement factors (R-values) of all the structures are in the range 0.17 to 0.18 with expected Ê , as estimated from coordinates error of 0.2 A Luzzati plots (Luzzati, 1952; Table 1). The stereochemistry of all the models is good, as judged by the PROCHECK calculation (Laskowski et al., 1993; Table 1), which marks over 98% of the residues as residing in most-favoured and additional allowed regions of the Ramachandran diagram (Ramachandran & Sasisekharan, 1968). The nearly perfect isomorphism between the crystals of all the forms (Table 1) leads us to believe that the subtle variations between the models are free of bias due to crystal forces and therefore re¯ect genuine differences between the structures. Tertiary and quaternary structure EcorL is a dimer with a subunit molecular mass of 28,763 Da as determined by mass spectrometry (Young et al., 1995). Only 239 amino acid residues out of the expected 242 (Young et al., 1995) could be traced in the electron density maps. In all the models, the well-ordered heptasaccharide bound covalently to Asn17 (Shaanan et al., 1991) was re®ned (with ®nal average temperature factors of Ê 2 for the models of ECPRO, 34, 30, 34, 40 and 37 A ECGAL, ECNGAL, ECLAC and ECNAL, respectively). In ECNGAL a faint electron density could also be observed at the 0.5s level around the second glycosylation site on Asn113 (Young et al., 1995), probably because of the better quality of the diffraction data (Table 1). However, attempts to build a covalently bound carbohydrate molecule into the map and re®ne it, failed. Overall, the structure of the complexes and the free form of EcorL are very similar. The Ca atoms Ê from of the ®ve models lie within 0.075 to 0.12 A the average Ca positions, whereas the estimated standard deviations (e.s.ds) from the averaged positions for all the atoms are 0.274, 0.358, 0.429, Ê for ECPRO, ECGAL, ECNGAL, 0.256 and 0.300 A ECLAC and ECNAL, respectively. A few residues adopt different orientations (e.g. Tyr82 in ECGAL and Gln117 in ECNAL and ECLAC) without any obvious reason, but in line with the inherent ¯exibility of proteins. In ECNGAL, alternative conformations for several residues could be observed and re®ned. In ECGAL, both the a and b anomers of the Gal 1-OH were modelled and re®ned, with occupancies of 0.45 and 0.55 for the a and b anomers, respectively, giving the best results, as judged by the cleanliness of Fo ÿ Fc difference maps around the 1-OH group. In solution, the a/b ratio is 0.29/0.64 (Vyas et al., 1994), thus suggesting that in the ECGAL crystal the ratio tips somewhat in

919

Structures of the Erythrina corallodendron Lectin Table 1. Statistics of data collection and re®nement A. Data collection Total number of reflections Number of unique reflectionsa Completeness Ê) Resolution (A Rsym Rsym in last resolution shell Overall redundancy Redundancy in last shell Unit cell dimensions Ê) a (A Ê) b (A Ê) c (A b (deg.) B. Refinement Final R-factor Rfree Total number of reflections included in refinementb Total number of atomsc Number of water molecules Ê 2) Average B-factor (A Protein atoms All atoms Coordinates error (Luzzati plot)d R.M.S. deviations Ê) Bond (A Angle (deg.) Dihedral (deg.) Improper (deg.) Ê 2) Isotropic B-factor restraints (A Main-chain bond Main-chain angle Side-chain bond Side-chain angle Residues ine Most favoured regions(%)f Additional allowed regions(%) Generously allowed regions(%) Disallowed regions(%)g

ECPRO

ECGAL

68,264 29,424 0.95 1.95 0.062 0.240 2.3 2.0

99,253 28,798 0.95 1.95 0.060 0.243 3.4 3.3

84.605 72.881 71.338 113.48 0.177 0.200 27,037 (2152) 2079 142

ECNGAL

ECLAC

ECNAL

123,619 30,859 0.99 1.90 0.049 0.145 4.0 3.8

60,388 27,823 0.91 1.95 0.056 0.263 2.2 2.0

65,763 27,182 0.94 1.95 0.052 0.267 2.4 2.1

84.202 72.990 71.290 113.44

84.223 72.980 71.259 113.33

84.061 72.974 71.243 113.40

83.933 73.174 71.311 113.36

0.179 0.206 27,809 (1710) 2096 135

0.181 0.200 30,841 (939) 2166 167

0.174 0.200 26,067 (2583) 2108 148

0.169 0.177 26,190 (2062) 2117 155

25.7 27.4 0.21

21.6 23.2 0.20

23.2 25.1 0.20

30.7 32.4 0.22

25.5 27.5 0.21

0.007 1.5 27.5 1.3

0.005 1.4 27.1 1.2

0.005 1.4 26.9 1.2

0.007 1.5 27.1 1.3

0.006 1.5 27.2 1.2

1.8 2.7 3.7 5.4

1.5 2.2 2.6 3.9

1.4 2.1 2.4 3.7

1.5 2.4 3.3 4.8

1.6 2.5 3.3 4.9

90.1 9.4 0.5 0.0

89.7 9.8 0.5 0.0

89.7 9.8 0.5 0.0

88.7 10.8 0.5 0.0

89.2 10.3 0.5 0.0

a

No s cutoff was applied. Ê. In parentheses, number of re¯ections in test set (BruÈnger, 1992b). Lower resolution limit for re®nement was 6.0 A c Including atoms in alternative conformations in ECGAL and ECNGAL. d The coordinate errors according to the cross-validated Luzzati plot (Kleywegt & BruÈnger, 1996) are nearly identical. e As output by PROCHECK (Laskowski et al., 1993). f Non-glycine and non-proline. g Including glycine and non-proline. b

favour of the a anomer. It is not clear whether this apparent difference in ratio between the a/b anomers in solution and in the crystal is due to a selection by the lectin or merely re¯ects lack of full atomic resolution. Unlike in the Gal/Glc-periplasmic transport protein, where clear selection of the b anomer is observed in the crystals because of preferential interaction with the protein (Vyas et al., 1994), the 1-OH group in either anomer of Gal is not involved in close contacts or hydrogen bonds with protein atoms in ECGAL. However, the 1-OH group in the a anomer makes two hydrogen bonds Ê ) and 592 (3.18 A Ê ), to water molecules 625 (3.14 A whereas in the b anomer the 1-OH group makes only one hydrogen bond to water molecule 625 Ê ; Figure 3b). (2.95 A Legume lectins are oligomers consisting of either two or four subunits. The subunit consists of two

b-sheets, with interconnecting loops. The similarity between the three-dimensional structure of the subunits of various legume lectins has been noted in earlier studies and is independent of the carbohydrate speci®city of the lectin (Einspahr et al., 1986; Bourne et al., 1990a,b; Shaanan et al., 1991; Delbaere et al., 1993; Loris et al., 1994; Banerjee et al., 1996). As could be expected, the similarity is particularly noticeable in the b-sheet regions. The inherently ¯exible interconnecting loops exhibit much larger variability. The ®rst few legume lectins to be determined crystallographically, namely, concanavlin A (Con A; Becker et al., 1975; Hardman & Ainsworth, 1972), PL (Einspahr et al., 1986), favin (Reeke & Becker, 1986) and LOL I (Bourne et al., 1990a), showed a recurrent dimeric motif, consisting of an extended 12-stranded b-sheet, which is formed

920 across the 2-fold axis relating the two subunits of the lectin dimer. This dimeric structure, currently known as the canonical dimer of the legume lectins (Shaanan et al., 1991), has since been found in other structures such as the lentil lectin (Loris et al., 1994) and SBA (Dessen et al., 1995). Since the canonical dimer is stabilized by a series of typical intra b-sheet hydrogen bonds across the dimer 2-fold axis and by hydrophobic forces, it was expected to prevail throughout the legume lectin family. However, as was ®rst found in EcorL, alternative, non-canonical, quaternary structures can exist among legume lectins (Shaanan et al., 1991). In EcorL, the breakdown of the canonical dimer is most likely caused by the covalently bound carbohydrate molecule on Asn17, which drives the two monomers apart, leading to the formation of an alternative dimer interface (Shaanan et al., 1991). Other legume lectins that have been determined more recently (GS4 and PNA) assemble to different types of non-canonical dimers without any effect of covalently bound carbohydrate (Delbaere et al., 1993; Banerjee et al., 1996). Since the function of legume lectins in their natural environment is not known (Weis & Drickamer, 1996; Rini, 1995), the question of the relation between the quaternary structure and their function remains unanswered at this stage. If however, the strong tendency of legume lectins to cross-link complex carbohydrate structures is indicative of their function in vivo, it is conceivable that such a relation exists, since the quaternary structure obviously determines the mutual spatial

Structures of the Erythrina corallodendron Lectin Ê) Table 2. Direct and indirect hydrogen bonds (<3.5 A between the two monomers at the ECNAL dimer interface MonA O P186-Wat 605 Wat 605-Wat 604 Wat 604 O G189 O I191 O I191 N I191-Wat 603 Wat 603 OG1 T193 OG1 T193-Wat 606 Wat 606a N T193-Wat 544 Wat 544 NE2 Q156-Wat 543 O T193-Wat 543 Wat 543 NE2 H180 OD1 D173

MonB

Wat 584 Wat 584-ND2 N167 Wat 584 Wat 584-ND2 N167 NH1 R73 NH2 R73 NH2 R73 ND1 H180 OG1 T193 Wat 654 Wat 654-OE1 E71 NZ K171 Wat 542 Wat 542 Wat 542-N A195 Wat 542-NZ K154

Ê) Distance (A 2.74 2.77 2.85 2.76 2.94 2.76 3.10 3.17 3.12 2.82 2.94 3.06 3.14 2.96 2.80 3.05 2.79 2.80 3.48 3.37 2.76 2.94 3.30

The names of atoms participating in the hydrogen bonds per monomer are given, along with the hydrogen bond distances. The complete set of hydrogen bonds at the dimer interface would be generated by applying the 2-fold operation. a Water molecule 606 is on the dimer 2-fold axis (Figure 1).

disposition of the combining sites in the lectin oligomer. The unusual interface of EcorL was observed in two crystal forms (Shaanan et al., 1991) and thus

Figure 1. The dimer interface in EcorL (stereo). The two EcorL monomers forming the dimer are related by the crystallographic 2-fold axis at 0,y,1/2, which is slightly tilted and roughly in the plane of the Figure. Residues emanating from b-strands of each monomer interdigitate and form direct or water-mediated hydrogen bonds, as well as van der Waals contacts (see also Table 2 and Figure 2). Monomer A, strands in green, side-chains and water molecules, depicted as spheres, in cyan; monomer B, strands in red, side-chains and water molecules in yellow; residues are labelled in the colour of the corresponding monomer. Water molecule 606 on the 2-fold axis is depicted as a large orange sphere.

Structures of the Erythrina corallodendron Lectin

undoubtedly represents the genuine quaternary structure of this lectin. In this interface, the two monomers interact in a ``hand-shake'' mode, reminiscent of the interface between VH and VL domains encountered in Fab structures. The EcorL dimer interface is held by three direct and 11 water-mediated hydrogen bonds per monomer (Table 2, Figure 1), as well as by van der Waals interactions (113 inter-monomer distances shorter Ê were found at the interface). The total than 4.0 A solvent-excluded area between the monomers is Ê 2 (see Materials and Methods). Overall, the 2336 A ®ne details of the dimer interface persist among all the ®ve models with the exception of the better resolved ECNGAL, where two additional hydrogen bonds are observed. The additional hydrogen bonds in ECNGAL are: O Ala195 (mon A)Wat536-O Ser176 (mon B) and OG Ser175 (mon A)-OE1 Glu196 (mon B). It is interesting to note that a completely buried water molecule is trapped in the middle of the interface on the 2-fold axis between His180 and Thr193 of the two monomers and is hydrogen bonded to both symmetry-related Thr193 residues (Figure 1, Table 2). The combining site Carbohydrate-binding proteins have been classi®ed according to the solvent exposure of their combining sites (Quiocho et al. 1989; Vyas, 1991; Rini, 1995). Group I consists of proteins with buried sites, such as enzymes and the bacterial periplasmic transport proteins (Quiocho, 1989; Zou et al., 1993), while group II consists of lectins with exposed sites. As in other legume lectins, the combining site of EcorL is a depression on the surface of the protein. Several residues that are crucial for ligand binding (Ala88, Asp89, Gly107 and Asn133) are absolutely conserved among all legume lectins regardless of their carbohydrate speci®city (Sharma & Surolia, 1997). Together with the conserved aromatic residue (Phe131 in EcorL), these residues retain identical spatial disposition within the combining sites, probably because of the proximity to the conserved Ca2‡-binding site (Shaanan et al., 1991; Rini, 1995). If the terminology of antibodies is borrowed, these residues can be considered as framework residues for ligand binding within the legume lectin family, whereas variable residues determine the particular speci®city of each lectin (Shaanan et al., 1991; Reeke & Becker, 1988; Rini, 1995). A stretch of variable residues essential for determining the speci®city toward galactose is spanned in EcorL between residues 217 and 223 (Shaanan et al., 1991; this stretch is part of loop D of the legume lectins, according to a recent naming convention; Sharma & Surolia, 1997). The complementarity between the electrostatic potential at the combining site of EcorL and the ligand (Gal) is apparent in Figure 2. The potential was calculated and depicted using the programs Delphi and Grasp (Nicholls & Honig, 1991;

921

Figure 2. Representation of the electrostatic potential of the EcorL combining site and of galactose. The potential map was calculated and depicted using the programs Delphi and Grasp (see Materials and Methods). For convenient representation, the surfaces covering the combining site residues and the ligand were separated and rotated by 45 around a vertical axis in the plane of the Figure. Surfaces are coloured such that red is for negative potential (ÿ2kt/e) and blue for positive potential (‡2kt/e), changing gradually through white. In this view, Gal is viewed roughly down the B-face. Note the large positive patch on this face (see the text). The location of residues and atoms in the combining site and in Gal are marked.

Nicholls et al., 1991; and see Materials and Methods). A prominent feature of the potential in the combining site is a large negative spot around OD1 and OD2 of Asp89, which accepts the positive patch on the edge of Gal between O3 and O4 (Figure 2). The large positive patch on the B-face of Gal seen in Figure 2 is due to the partial positive charge on the hydrogen atoms of C3, C4, C5 and C6 (Weis & Drickamer, 1996). This positive region ®ts under Phe131 to form the aromatic stacking that is a prominent feature in Gal-speci®c lectins (Rini, 1995; Weis & Drickamer, 1996; Elgavish & Shaanan, 1997). The partial negative charge on O3 Gal ®ts into the positive depression surrounding ND2 Asn133 and N Gly107, while the partial negative charge of O4 and O6, due to their lone electron pairs, ®ts with the partial positive charge around N Ala218 and NE2 Gln219, respectively. Thus, it seems as if the electrostatic potential plays an important role in positioning the ligand in the primary binding site of EcorL, along with the requirements to optimize the hydrogen bonding and van der Waals interactions with the protein.

922

Structures of the Erythrina corallodendron Lectin

An important question in studies on ligand binding to proteins (or substrate binding to enzymes) is that of the conformational changes associated with the binding. The combining sites of lectins appear to be generally preformed (Weis & Drickamer, 1996). However, some variations on this theme emerge when the new structural information on EcorL is put together with the information available on other legume lectins. On the basis of a comparison between the unliganded and liganded states of PL, Rini et al. (1993) identi®ed a conformational change due to ligand binding in the loop encompassing residues 216 to 218 (equivalent to the speci®city loop D of EcorL), which is manifested in a movement of backbone Ê between the two forms. atoms of up to 1.1 A Support for the validity of this observation comes from the fact that a similar, although smaller, change has been observed in LOL I by Bourne et al. (1990b) and in the lentil lectin by Loris et al. (1994), structures that have been determined at a higher resolution than the PL-trimannoside (Rini et al., 1993; see below). Furthermore, on the basis of superposition of the liganded PL on the unliganded Con A structure (Hardman et al., 1984), Rini et al. (1993) suggested, with reservations, that Con A may not undergo conformational change upon ligand binding, thus indicating perhaps that its combining site is more preformed to accept the entering ligand than that of PL. Interestingly, these authors noted also that two water molecules found

in the combining site of unliganded Con A occupy positions identical with those taken by the 4-OH and 6-OH groups of the Man bound to PL , but that water molecules in the combining site of unliganded PL (Prasthoffer et al., 1989), are somewhat removed from the positions occupied by the sugar hydroxyl groups in the complex (after superposition of unliganded PL on the liganded PL, PDB codes 2ltn and 1rin, respectively, the following Ê, distances were measured: 3-OH-Wat 559: 0.87 A Ê , O5-Wat 562: 0.94 A Ê and 4-OH-Wat 494: 0.59 A Ê ). Although the somewhat O6-Wat 561: 1.35 A inferior resolution at which the PL-trimannoside Ê ; Rini et al., structure has been determined (2.6 A 1993) may cast doubt on the signi®cance attributed to these differences, it should be noted that similar differences have been observed between the location of water molecules in the combining site of the unliganded LOL I structure determined at Ê (Bourne et al., 1990a) and the hydroxyl 1.9 A groups of the bound ligands, as observed in the Ê complexes with Man and Glc determined at 1.9 A Ê and 2.2 A, respectively (Bourne et al., 1990b). In ECPRO, the uliganded state of EcorL, two well-de®ned water molecules, 589 and 609, are found in the exact positions inhabited by the 3-OH and 4-OH groups, respectively, of the Gal bound at the primary binding site. These water molecules maintain the same hydrogen bonding pattern with the framework residues stabilizing the ligand (Figures 3 and 4; Table 3). At the same time, no

Figure 3a (legend on page 924)

Structures of the Erythrina corallodendron Lectin

Figure 3b ± c (legend on page 924)

923

Figure 3. Schematic, two-dimensional diagrams (LIGPLOT; Wallace et al.,1995) of the combining sites of EcorL and its complexes. Carbon atoms are depicted in black, oxygen in red, nitrogen in blue, hydrogen bonds (broken lines) and their length in green. The water molecules are depicted as spheres in colours corresponding to the colour scheme in Figure 4. a, ECPRO; b, ECGAL (only the b-anomer of Gal is depicted); c, ECGNAL; d, ECLAC; e, ECNAL.

925

Structures of the Erythrina corallodendron Lectin

Figure 4. Distribution of water molecules in the combining sites of EcorL and its complexes (stereo). The models were superimposed (see Materials and Methods) and the water molecules in combining sites were depicted as spheres (pink for ECPRO, green for ECGAL, orange for ECNGAL, purple for ECLAC and cyan for ECNAL). For clarity, only the side-chains in the combining site of ECNAL are shown and labelled (white). N-Acetyllactosamine (red stick bonds) and GalNac (orange bonds) represent the ligands, and the C3, C4 and C5 atoms of Gal are labelled (red). Clusters of water molecules are numbered (yellow, italics) according to Table 3. Labels of isolated water molecules are coloured according to the colour scheme above. Note that water molecules 609 and 589 of ECPRO superimpose exactly on the positions of O3 and O4, respectively, of Gal.

conformational change is observed in the protein between ECPRO and any of the complexes. A similar situation is encountered in GS4, where water molecules 437 and 438 of the unliganded form

reside in the exact positions of the 3-OH and 4-OH groups, respectively, of the Gal moiety of the bound Lewis b (Delbaere et al., 1993). Thus, it seems as if the more re®ned picture emerging from

Table 3. Array of water molecules in the combining sites of EcorL and its complexes Cluster no.

ECPRO

ECGAL

ECNGAL

ECLAC

ECNAL

1

527 (20)

2

527 (25) l 546 (40)

522 (31) l 523 (40)

527 (25) l 528 (35)

3

553 (45) l 565a (58)

527 (33) l 549 (36)

524 (52)

529 (44)

4 5

595 (55)

626 (55) l 627 (62) l 525 (59) l 526 (55) 527 (25) 588 (60)

639 (54) l 640 (73) l 530 (42) l 531 (58) 532 (66) 597 (42)

528 (53) 10

9

592 (54) l 625 (54)

613 (55)

6 7

642 (42)

8 9 10a 10b

583 (75)

579b (59)

606 (59)

595 (65) 585c (51)

6

7

535(57) l 589 (23) l 609 (22) Total

7

Water molecules in all ®ve models after superposition, as described in Materials and Methods, were divided into clusters. Numbering of clusters refers to Figure 4. Numbering of individual water molecules within the cluster refers to the numbering in the PDB ®le of each model. Isolated water molecules are numbered as in the PDB ®le and as in Figure 4. The B-factor for each water molecule is in parentheses. Hydrogen bonds between water molecules are marked with l. For full details on hydrogen bonding consult Figure 3. a Ê away from cluster 3 (Figure 4). Water 565 in ECPRO is about 2 A b Ê away from cluster 9 (Figure 4). Water 579 in ECNGAL is about 2 A c Ê ). Hydrogen bonded to Tyr108 OH (2.93 A

926 the above observations on the structures of legume lectins and their complexes determined so far, is that the Man (or Glc) speci®c lectins, with the possible exception of Con A, tend to undergo a small conformational change in the D loop upon ligand binding, whereas the conformation of the galactose-speci®c lectins (GS4 and EcorL) remains ®xed. Furthermore, the fact that in the free forms of GS4 and EcorL, water molecules are located exactly in the positions to be occupied by hydroxyl groups of the incoming ligand, suggests that already in the uliganded state, the potential ®eld in their combining sites is probably tuned for accepting the Gal 3-OH and 4-OH hydroxyl groups. On the other hand, the potential ®eld in the combining sites of PL and related lectins has to be adjusted in order to optimize the hydrogen bonding pattern with Man (or Glc). The conformational changes accompanying the binding to PL and the other Glc/Man-speci®c lectins mentioned above, may well be driven, at least in part, by the requirement for such an adjustment. However, whereas the currently available information might indicate some correspondence between the lack of conformational change upon ligand binding and the presence of water molecules in the exact location of hydroxyl groups of the incoming ligand, further structural work is required in order to establish whether such correspondence holds in general and whether it is related in any way to the lectin speci®city. Binding of monosaccharides to EcorL In all the four complexes of EcorL studied here, the galactose moiety in the primary site is bound in an exactly identical way. This is not surprising, since the galactose molecule is held ®rmly by a series of hydrogen bonds with side-chain and main-chain atoms of the framework residues Asp89, Gly107 and Asn133, and of the variable loop residues Ala 218 and Gln219, as well as by van der Waals interactions with the conserved residues Ala88 and Phe131 (Shaanan et al., 1991). In addition, a strong water-mediated hydrogen bond between O6 and O Leu86 seems to play an important role in the binding (Figure 3). The tight grip of EcorL on the Gal moiety is manifested in the relatively low mobility of the Gal moiety, which is Ê 2 below the overall temperature factor of about 3 A the model in all four complexes. The binding of N-acetylgalactosamine (GalNac) in ECNGAL does not involve any new direct hydrogen bonds with the protein compared to galactose. Rather, two additional water-mediated hydrogen bonds to ND2 Asn133 and OH Tyr108 (Figures 3 and 4; Table 3), as well as additional van der Waals contacts can be observed in this complex. Binding of disaccharides to EcorL Binding of a second monosaccharide molecule (Glc in ECLAC and GlcNac in ECNAL) involves

Structures of the Erythrina corallodendron Lectin

more van der Waals contacts but addition of only one direct protein-carbohydrate hydrogen bond in ECLAC and two in ECNAL (Figure 3). Thus, in both ECLAC and ECNAL, a hydrogen bond is formed between NE2 Gln219 and O3 of Glc or GlcNac, with an additional, probably weaker, hydrogen bond between NE2 Gln219 and O7 of the acetamido moiety of GlcNac (Figure 3). It is interesting to note that as more hydrogen bonds between Gln219 and the ligand are formed, the relative mobility of the side-chain of this residue is reduced, as manifested by the difference between the average temperature factors of the side-chain and the overall temperature factors of the models. In ECPRO, the unliganded state, the difference is Ê 2, in ECGAL and ECNGAL the differences 24.2 A Ê 2, respectively, whereas in are 16.3 and 16.6 A Ê 2, ECLAC and ECNAL they are 7.5 and 4.4 A respectively. A marginal reduction in the relative mobility of GlcNac in ECNAL, possibly resulting from the additional hydrogen bond between NE2 Gln219 and O7, is observed (the difference between overall temperature factor and the average temÊ 2 for Glc in ECLAC and perature factor is 19.8 A Ê 2 for GlcNac in ECNAL). Curiously, the 14.5 A reduction in the mobility of Gln219 upon progression from the unliganded state to the disaccharidic complexes is manifested also in the relative temperature factor of a water molecule hydrogen bonded to OE1 Gln219 on the external side of the combining site in all the complexes (Figure 3; the differences between the temperature factor of the Gln219 bound water and the overall temperature factor are 18.0, 17.6, 19.3, 12.1 and Ê 2 in ECPRO, ECGAL, ECNGAL, ECLAC and 7.3 A ECNAL, respectively). It can also be noticed that the hydrogen bond NE2 Gln219± O6 Gal in the complexes with the monosaccharides is somewhat shorter than it is in the complexes with the disaccharides (2.95, 2.81, Ê in ECGAL, ECNGAL, ECLAC and 3.36 and 3.20 A ECNAL, respectively; Figure 3) and re¯ects a slight shift of the Gln219 side-chain in the two latter Ê difference between the complexes. This shift (0.5 A mean position of NE2 Gln219 in ECLAC and ECNAL and the mean position in ECGAL and ECNGAL) is probably necessary in order to optimize the stereochemistry of the additional hydrogen bond with O3 Glc (or GlcNac; Figure 3). Further accommodation of the acetamido moiety in ECNAL is achieved by a small change in the f/c dihedral angles around the b1 ! 4 glycosidic bond of the N-acetyllactosamine moiety (ÿ79.6 / ÿ139.7 ) compared to lactose (ÿ76.5 /ÿ143.2 ; the f/c angles are de®ned as O5-C1-O1-C40 and C1-O1-C40 -C50 , respectively) and not through a change in the conformation of the side-chain of Gln219, as suggested by modelling (Moreno et al., 1997). The values of the f/c angles observed in ECLAC fall within the range of f/c angles spanned by the glycosidic bond in 90% of the lactose molecules in water, as found in solution NMR measuremnts supported by force-®eld calcu-

Structures of the Erythrina corallodendron Lectin

lations (Asensio & Jimenez-Barbero, 1994; Espinosa et al., 1996). Binding of disaccharides via the same set of residues that are involved in binding monosaccharides at the primary site is not unique to EcorL and stands in contrast to the situation in Con A, where residues outside the primary site are recruited for binding additional monosaccharides (Naismith & Field, 1996). This difference may re¯ect the presence of the longer variable D loop encompassing Ala218 and Gln219 in EcorL (Shaanan et al., 1991; Sharma & Surolia, 1997). Examples of lectins in which the same residue binds the monosaccharide in the primary site and the second monosaccharide are galectin-1 and galectin-2 (Liao et al., 1994; Lobsanov et al., 1993) and the legume lectin PNA (Banerjee et al., 1996). In the complex of PNA and lactose, Ser211 on the D loop forms hydrogen

927 bonds with both O3 Glc and O4 Gal in the primary site, but a stronger hydrogen bond to the O3 Glc is provided by N Gly213, which is not involved in binding Gal in the primary site (Banerjee et al., 1996). The D loop in PNA is somewhat shorter than in EcorL, a long residue equivalent to Gln219 of EcorL is not involved in Glc binding, and therefore the net result is that the Glc moiety of the bound lactose in PNA is pulled closer to the main chain of this loop than in EcorL (see Figure 7b of Banerjee et al., 1996). Consequently, the conformations of the glycosidic bonds of the lactose in PNA and EcorL are somewhat different (Figure 5b and c). The critical role played by O3 of Glc (or GlcNac) in binding the second monosaccharide to lectins is the common theme emerging from comparison of the complexes with lactose and N-acetyllactosa-

Figure 5. Distribution of hydrogen bond partners around the second saccharide in complexes with lactose and N-acetyllactosamine (stereo). Atoms forming hydrogen bonds with the carbohydrate in the extended site and names of PDB ®les are given in Table 4 and are depicted as spheres, carbohydrate in stick bonds and water molecules as small spheres. Hydrogen-bonded carbohydrate atoms are labelled. a, Superposition on atoms of the hexose ring of GlcNac of the three N-acetyllactosamine molecules from the complexes of EcorL (ECNAL, yellow), galectin-1 (cyan) and soybean agglutinin (green); b, superposition on atoms of the hexose ring of Glc of ®ve lactose molecules from complexes of EcorL (ECLAC, yellow), peanut lectin (white), heat-labile enterotoxin (green), ricin (cyan) and galcetin-2 (red); c, The same as b, but superposition on the Gal moiety.

928

Structures of the Erythrina corallodendron Lectin Table 4. Hydrogen bonds with the second saccharide in complexes of lectins with lactose and N-acetyllactosamine Proteina

O3b

N2

NH2 Arg73 NH1 Arg48 OE2 Glu71 OD1 Asp215 NE2 Gln219

Wat6d

Wat597

O3

O2 NH2 Arg70

Ricin (2aai)h LT (1ltt)i

NH2 Arg70 NH2 Arg49 OE1 Glu68 OG Ser211 N Gly213 OD1 Asp25 O Gln56

EcorL (ECLAC)

NE2 Gln219

Wat588

A. N-Acetyllactosamine Galectin-1 (1slt)c SBA (1sba)e EcorL (ECNAL) Protein B. Lactose Galectin-2 (1hlc)f Peanut lectin (2pel)g

Wat189

O6

O7

Wat640

NE2 Gln219 O6

Wat681 Wat259 Wat146 Wat124 Wat627

a

Shortened and full names of proteins as listed in the PDB ®le and the PDB code: SBA, soybean agglutinin; LT, heat labile enterotoxin. b Atoms of Glc or Gal NAc involved in hydrogen bonds with the protein atoms below. c Liao et al. (1994). d All water molecules in the Table are also hydrogen bonded to protein atoms except for ricin. e Dessen et al. (1995). f Lobsanov et al. (1993). g Banerjee et al. (1996). h Rutenber & Robertus (1991). i Sixma et al. (1992).

mine, whose coordinates are available in the PDB (Table 4; Figure 5). According to the information currently available, in all the eight complexes listed in Table 4, the O3 atom participates in hydrogen bond with the protein. In the complexes with lactose, further hydrogen bonding is provided via O6 and occasionally, through O2 (Table 4; Figure 5b and c). In the three known complexes of N-acetyllactosamine, N2 seems to contribute to the stabilization through hydrogen bonding of the GlcNac moiety. Since O6 and O2 of Glc and N2 of GlcNac are usually more exposed to the solvent than O3, the hydrogen bonds with them are usually water-mediated, thus demonstrating again the role of water molecules in reaching out towards carbohydrate atoms that are not in direct contact with the combining site (Weis & Drickamer, 1996). The f/c angles for the lactose complexes given in Table 4 are in the range ÿ76.5 to ÿ89.4 for f and ÿ123.5 to ÿ143.2 for c. The range of the f/c angles for the complexes with N-acetyllactosamine is ÿ80.1 to ÿ95.1 for f and ÿ101.4 to ÿ137.0 for c. Since O3 of Glc (or GlcNac) appears to be a dominant factor in stablizing the second saccharide, the conformation of the glycosidic bond seems to be governed by the need to satisfy the constraints imposed by the protein atoms hydrogen bonding O3.

Reorganization of water molecules in the combining site Reorganization of water molecules in the combining sites as a result of ligand binding has been observed in several lectins (Weis & Drickamer, 1996). The contribution of solvent reorganization to the thermodynamics of ligand binding to proteins and lectins has been emphasized by Chervenak & Toone (1994) and could be a signi®cant source for the subtle differences observed in the binding process between lectins with overlapping af®nity (Toone, 1994; Chervenak & Toone, 1995). The distribution of water molecules in the ®ve EcorL structures is depicted in Figure 4 and Table 3. Superposition of the combining sites reveals that while several water molecules retain nearly identical positions in all the structures, others change positions depending on the particular ligand in the combining site. For convenient analysis of the solvent reorganization, water molecules that are close in space in all the structures were grouped into clusters. As can be seen in Figure 4, clusters 1 and 2 represent two groups of water molecules present in all the structures. Cluster 1 consists of well-de®ned water molecules that are involved in tethering the 6-OH group of Gal to O Leu86. The tight hydrogen bonds of these water molecules to both 6-OH and

929

Structures of the Erythrina corallodendron Lectin

O Leu86 (Figure 3; Table 3), prompts us to suggest that they could be considered as an intrinsic part of the combining site and as an extension of the protein matrix facing the entering ligand (Toone, 1994). Clusters 4 and 10b consist of water molecules in ECGAL and ECNGAL, respectively, that would be displaced by Glc or GlcNac in the complexes with disaccharides. Water 579 of ECNGAL, which is somewhat removed from cluster 9, is displaced by the acetamido group of GalNac and is involved in a water-mediated hydrogen bond through water 585 to OH Tyr108 (Figures 3 and 4; Table 3). Water 528 of ECLAC, which is hydrogen bonded to N Gly107, is absent from all the other complexes, but only in ECNGAL can its absence be rationalized on the basis of steric clash with the acetamido group of GalNac. Overall, water molecules of ECLAC and ECNAL are found in 9 out of the 11 clusters (Table 3; Figure 4), an observation that supports the notion that each additional saccharide nucleates further ordering of water molecules in the combining site, although not necessarily in its immediate vicinity. Prominent examples of water ordering in lectin-carbohydrate complexes are the complexes between LOL I and oligosaccharides (Bourne et al.,1992) and between GS4 and Lewis b tetrasaccharide (Delbaere et al., 1993). Interestingly, an increase by two or three additional water molecules upon binding of lactose and Nacetyllactosamine to EcorL as compared with that of galactose has been reported (Surolia et al., 1996). It should be borne in mind however, that only the less mobile water molecules are observed in the electron density maps and hence are included in the models (Levitt & Park, 1993). Therefore, the full extent of the water reorganization as a result of binding to proteins in general and to the complexes of EcorL in particular, can only be estimated on the basis of crystallographic studies. Moreover, since the crystals of EcorL and of its complexes were grown from 2,4-methylpentanediol (MPD), which in itself can affect the water structure around the protein, further care should be applied in interpreting the binding data (see below) in light of the evidence concerning reorganization of water molecules in the combining site.

Correlation with thermodynamic data on binding to EcorL The binding data for the four complexes as obtained by Surolia et al. (1996) are summarized in Table 5, together with data emerging from our structural work. Along the lines of the analysis presented by Toone (1994) and bearing in mind his cautionary remarks about correlating binding data to lectins with structural information, the present study enables us to draw a few qualitative conclusions, mostly of comparative nature, about the binding parameters of the various complexes. The formation of EcorL complexes is characterized by enthalpy-entropy compensation (Toone, 1994; Surolia et al., 1996). Table 5 shows that the binding enthalpy increases with the amount of area buried upon complex formation. As pointed out by Toone (1994) and others (see references in Toone, 1994), burial of surfaces would have an enthalpic effect due to reorganization of the solvent surrounding the surfaces of the carbohydrate and protein (Beierbeck et al., 1994). Apart from the increase in buried area, the rise in binding enthalpy upon moving from monosaccharidic (ECGAL and ECNGAL) to disaccharidic complexes, correlates with the addition of direct hydrogen bonds between the protein and the ligand (one in ECLAC: O3 Glc-NE2 Gln219 and two in ECNAL: O3 GlcNac-NE2 Gln219 and O7 GlcNac-NE2 Gln219; Figure 3). Another source of the increased binding enthalpy for the disaccharides are the more extensive van der Waals interactions with the protein. Inspection of Table 5 reveals that in contrast to binding of the other ligands, binding of Gal to EcorL is accompanied by an increase in entropy. A possible rationalization for this phenomenon is that in the case of Gal binding, the entropy increase due to release of the tightly bound water molecules 589 and 609 (Figures 3 and 4; Table 3), may still not be fully compensated by opposing factors contributing to decrease in entropy. Thus, Gln219 undergoes only a small reduction in mobility in ECGAL (see above). Furthermore, the number of detectable water molecules in the combining site of ECGAL is about the same as that in ECPRO, which suggests that there may not be an apparent increase in the ordering of water mol-

Table 5. Information on binding to EcorL complexes

Model ECGAL ECNGAL ECLAC ECNAL ECPRO

Kb (10 l molÿ1)

ÿ
Gb (kJ molÿ1)

ÿ
Hb (kJ molÿ1)

ÿ
Sb (J molÿ1 Kÿ1)

Buried area Ê 2) (A

1.57  0.13 1.34  0.11 1.94  0.12 9.73  0.6

18.2  0.2 17.9  0.2 18.8  0.1 22.7  0.1

13.7  1.0 23.0  1.6 41.2  2.5 47.1  2.6

ÿ15.3  3.4 17.1  5.3 75.4  8.4 83.2  9.5

455 561 616 643

3

Binding energies and constants taken from Surolia et al. (1996) for measurements around 298 K.

Number of water molecules in binding site 6 7 10 9 7

930 ecules in the combining site of ECGAL compared to the unliganded state (Table 3; Figure 4). It should be pointed out that the mobility of the sidechains of Asp89 and Asn133, which also are involved in direct hydrogen bonds with the ligand, hardly changes between ECPRO and the complexes. In all the models, the temperature factors of the Asp89 and Asn133 side-chains are, on average, Ê 2, respectively, below the overall tem7.5 and 5 A perature factor of the model, which probably re¯ects the fact that these side-chains are anchored by the neighbouring calcium ion, either directly (Asn133), or through a water molecule (Asp89). Thus, freezing of these two side-chains may not contribute signi®cantly to the con®gurational entropy of binding to EcorL (Toone, 1994). The decrease in entropy upon moving from ECGAL to ECNAL correlates with the trend of decreased Gln219 mobility mentioned above, localization of more water molecules and, to a certain extent, with the restraints imposed upon the conformational freedom around the glycosidic bond in ECLAC and ECNAL. Fixing the acetamido group of GalNac in ECNGAL has a similar effect. Computational studies, taking as a starting point the new structural information presented on the unliganded state of EcorL and on the complexes with mono- and disaccharides, should yield a more quantitative correlation with the thermodynamic data on binding to this lectin.

Materials and Methods Crystallization EcorL was isolated and puri®ed according to Lis et al. (1985) in the laboratory of Professor N. Sharon. Crystals were grown by the vapour diffusion method at room temperature using the hanging or sitting drop methods. Drops of 6 mg/ml protein, 15% (v/v) 2,4-methylpentane-diol (MPD, Fluka), 50 mM Hepes buffer (pH 7.0), 0.02% (w/v) sodium azide were equilibrated against reservoirs containing 30% MPD, 100 mM Hepes (pH 7.0), 0.02% sodium azide. For crystallization of the complexes, the particular ligand was added to a concentration of 10 to 15 mM in the drop prior to setting up the crystallization experiment (galactose and lactose were purchased from Fluka; N-acetyllactosamine was from Calbiochem; GalNac was a gift from Professor N. Sharon). It is possible that the presence of MPD favours formation of the complexes, since on the basis of the relatively low binding constants (Table 5), higher concentrations of ligands should be required in order for crystals with full ligand occupancy at the combining site to form. However, concentrations of ligands higher than 10 to 15 mM prevented formation of crystals, at least during the period of six to eight months during which the crystallization experiments were monitored. Typically, crystals reached a size of 0.8 mm 0.5 mm 0.5 mm within two weeks. The crystal data of each of the ®ve forms (unliganded EcorL and four complexes) are given in Table 1. The crystals contain one monomer in the asymmetric unit.

Structures of the Erythrina corallodendron Lectin Data collection Diffraction data were recorded at room temperature using a Rigaku (Japan) R-axis II detector mounted on a Rigaku RU-300 generator. The crystals were stable enough in the X-ray beam to enable recording a full data set using one crystal. Data were processed using Denzo, Scalepack (Otwinowski, 1993) and CCP4 programs (SERC, 1994). Refinement The starting model for ECLAC, ECNAL and ECPRO was the ECLAC model described by Shaanan et al. (1991). The initial model for ECGAL and ECNGAL was ECPRO. The initial models included the covalently bound heptasaccharide at Asn17. Re®nement was carried out using the programs X-PLOR (BruÈnger, 1992a; Adams et al., 1997) and CNS (Crystallography and NMR System, program under development, A.T. BruÈnger, personal communication). A similar procedure was used for re®ning all the models. For the purpose of cross-validation, re¯ections were selected for the test set prior to initiating the re®nement (BruÈnger, 1992b). Extensive simulated annealing starting from 3000 K was applied to each model in order to minimize model bias between the structures (Kleywegt & BruÈnger, 1996). Rigid body minimization was performed against data in shells of increasÊ and ending with 3 ing resolution, starting from 5 to 8 A Ê . This was followed by positional minimization to 6 A and simulated annealing (slow-cooling procedure) at 3000 K with a gradual increase of the high-resolution Ê (1.90 A Ê for ECNlimit of the data from 2.6 to 1.95 A GAL). The lower resolution limit of the data for the posÊ. itional and simulated annealing re®nement was 6.0 A All the re¯ections measured in these ranges were included throughout the re®nement, without applying any cutoff on the basis of sF/F ratio or magnitude. Models were modi®ed using the program O (Jones et al., 1991), after inspection of their ®t to 3Fo ÿ 2Fc, 2Fo ÿ Fc and Fo ÿ Fc maps calculated in X-PLOR. The appropriate ligands were ®tted into the maps at this stage. In the case of ECPRO, an anneal omit map (3000 K) in which a Ê around Asp89 in the region spanning a radius of 8 A combining site was calculated, in order to remove any model bias. After further re®nement, including minimization of individual B-factors, water molecules were added, following inspection of Fo ÿ Fc maps and occasionally by using the program ARP (Lamzin & Wilson, 1993). Re®nement proceeded at 600 K with harmonic restraints applied to the positions of the water molecules. Water molecules with individual B-factors Ê 2 were discarded from the model. In greater than 80 A the case of ECGAL, the a and b anomers of O1 were re®ned. The occupancy values for each anomer were veri®ed by inspection of the Fo ÿ Fc map. In ECNGAL, Ê resolwhich was re®ned against data extending to 1.9 A ution, alternative conformations were re®ned for several residues according to X-PLOR scripts. The ®nal statistics on the re®nement and quality of models are presented in Table 1. The coordinates and structure factors of the ®ve EcorL models have been deposited with the Brookhaven Protein Data Bank for immediate release (the codes for ECPRO, ECGAL, ECNGAL, ECLAC and ECNAL are 1axy,1axz, 1ax0, 1ax1 and 1ax2, respectively). For the purpose of comparison, models of EcorL and the complexes were superimposed on Ca atoms after omitting ten residues from the N and C termini. Superposition was carried out using X-PLOR or InsightII

Structures of the Erythrina corallodendron Lectin (Molecular Simulations) for the purpose of preparing Figures. The buried surface area was calculated with the CCP4 (SERC, 1994) programs AREAIMOL and DIFFARÊ. EA, using a probe radius of 1.4 A Delphi calculations (Nicholls & Honig, 1991) of the electrostatic potential were performed using the partial charges as in the Engh & Huber (1991) parameter set used in X-PLOR. Hydrogen atoms were included in the calculations. The sugar hydroxyl groups were rotated such that their hydrogen atoms pointed in the direction of the hydrogen bond acceptors. Figure 2 was made with Grasp (Nicholls et al., 1991). Figures 1, 4 and 5 were made using InsightII (Molecular Simulations). Figure 3 was generated by LIGPLOT (Wallace et al., 1995).

References Adams, P. D., Panu, N. S., Read, R. J. & BruÈnger, A. T. (1997). Cross-validated maximum likelihood enhances crystallographic simulated annealing re®nement. Proc. Natl Acad. Sci. USA, 94, 5018± 5023. Asensio, J. L. & Jimenez-Barbero, J. (1994). The use of AMBER force ®eld in conformational analysis of carbohydrate molecules: determination of the solution conformation of methyl a-lactoside by NMR spectroscopy, assisted by molecular mechanics and dynamics calculations. Biopolymers, 35, 55 ± 73. Banerjee, R., Das, K., Ravishankar, R., Suguna, K., Surolia, A. & Vijayan, M. (1996). Conformation, protein-carbohydrate interactions and novel subunit association in the re®ned structure of peanut lectinlactose complex. J. Mol. Biol. 259, 281 ± 296. Becker, J. W., Reeke, G. N., Wang, J. L., Cunningham, B. A. & Edelman, G. M. (1975). The covalent and three-dimensional structure of concanavalin A. J. Biol. Chem. 250, 1513± 1524. Beierbeck, H., Delbaere, L. T. J., Vandonselaar, M. & Lemieux, R. U. (1994). Molecular recognition. XIV. Monte Carlo simulation of the hydration of the combining site of a lectin. Can. J. Chem. 72, 463± 470. Bourne, Y., Abergel, C., Cambillau, C., Frey, M., RougeÂ, P. & Fontecilla-Camps, J. C. (1990a). X-ray crystal Ê structure determination and re®nement at 1.9 A resolution of isolectin I from the seeds of Lathyrus ochrus. J. Mol. Biol. 214, 571 ± 584. Bourne, Y., Roussel, A., Frey, M., RougeÂ, P., FontecillaCamps, J. C. & Cambillau, C. (1990b). Three-dimensional structures of complexes of Lathyrus ochrus isolectin I with glucose and mannose. Proteins: Struct. Funct. Genet. 8, 365 ± 376. Bourne, Y., RougeÂ, P. & Cambillau, C. (1992). X-ray structure of a biantennary octasaccharide-lectin Ê resolution. J. Biol. Chem. complex re®ned at 2.3 A 267, 197 ± 203. BruÈnger, A. T. (1992a). X-PLOR Manual Version 3.1, Yale University, New Haven, CT. BruÈnger, A. T. (1992b). Free R-value: a novel statistical quantity for assessing the accuracy of crystal structures. Nature, 355, 472 ± 475. Chervenak, M. C. & Toone, E. J. (1994). A direct measure of the contribution of solvent reorganization to the enthalpy of ligand binding. J. Am. Chem. Soc. 116, 10533± 10539. Chervenak, M. C. & Toone, E. J. (1995). Calorimetric analysis of the binding to lectins with overlapping

931 carbohydrate-binding ligand speci®cities. Biochemistry, 34, 5685± 5695. Delbaere, L. T. J., Vandonselaar, M., Prasad, L., Quail, J. W., Wilson, K. S. & Dauter, Z. (1993). Structure of the lectin IV of Griffonia simplicifolia and its complex with the Lewis b human blood group determined Ê resolution. J. Mol. Biol. 230, 950± 965. at 2.0 A Dessen, A., Gupta, D., Sabesan, S., Brewer, C. F. & Sacchettini, J. C. (1995). X-ray crystal structure of the soybean agglutinin cross-linked with a biantennary analog of the blood group I carbohydrate antigen. Biochemistry, 34, 4933± 4942. Einspahr, H., Parks, E. H., Suguna, K., Subramanian, E. & Suddath, F. L. (1986). The crystal structure of Ê resolution. J. Biol. Chem. 261, pea lectin at 3.0 A 16518± 16527. Elgavish, S. & Shaanan, B. (1997). Lectin-carbohydrate intractions: different folds, common recognition principles. Trends Biochem. Sci. 12, 462± 467. Engh, R. A. & Huber, R. (1991). Accurate bond and angle parameters for X-ray protein structure re®nement. Acta Crystallog. sect. A, 47, 392±400. Espinosa, J. F., Canada, F. J., Asensio, J. L., MartinPastor, M., Dietrich, H., Martin-Lomas, M., Schmidt, R. R. & Jimenez-Barbero, J. (1996). Experimental evidence of conformational difference between C-glycosides and O-glycosides in solution and in the protein bound state: the C-lactose/O-lactose case. J. Am. Chem. Soc. 118, 10862± 10871. Jones, T. A., Zou, J. Y., Cowan, S. W. & Kjeldgaard, M. (1991). Improved methods for building protein models in electron density maps and the location of errors in these models. Acta Crystallog. sect. A, 47, 110± 119. Hardman, K. D. & Ainsworth, C. F. (1972). Structure of Ê resolution. Biochemistry, 11, concanavalin A at 2.4 A 4910± 4919. Hardman, K. D., Agarwal, R. C. & Freiser, M. J. (1984). Manganese and calcium binding sites of concanavalin. A. J. Mol. Biol. 157, 69± 86. Kleywegt, G. J. & BruÈnger, A. T. (1996). Checking your imagination: applications of R value. Structure, 4, 897± 904. Lamzin, V. S. & Wilson, K. S. (1993). Automoated re®nement of protein models. Acta Crystallog. sect. D, 49, 129± 147. Laskowski, R. A., MacArtur, M. W., Moss, D. S. & Thornton, J. M. (1993). PROCHECK: a program to check the stereochemical quality of protein structures. J. Appl. Crystallog. 24, 946± 950. Levitt, M. & Park, B. H. (1993). Water: now you see it, now you don't. Structure, 1, 223± 226. Liao, D. L., Kapadia, G., Ahmed, H., Vasta, G. R. & Herzberg, O. (1994). Structure of S-lectin, a developmentally regulated vertebrate b-galactoside-binding protein. Proc. Natl Acad. Sci. USA, 91, 1428± 1432. Lis, H., Joubert, F. J. & Sharon, N. (1985). Isolation and properties of N-acetyllactosamine-speci®c lectins from nine Erythrina species. Phytochemistry, 24, 2803± 2809. Loris, R., van Overberge, D., Dao-Thi, M. H., Poortmans, F., Maene, N. & Wyns, L. (1994). Structural analysis of two crystal forms of lentil lecÊ resolution. Proteins: Struct. Funct. Genet. tin at 1.8 A 20, 330± 346. Lobsanov, Y. D., Gitt, M. A., Lef¯er, H., Barondes, S. H. & Rini, J. M. (1993). X-ray crystal structure of the human dimeric S-lac lectin, L-14-II, in complex

932

Structures of the Erythrina corallodendron Lectin

Ê reolution. J. Biol. Chem. 268, with lactose at 2.9 A 27034± 27038. Luzzati, V. (1952). Traitement statistique des erreurs dans la determination des structures cristallines. Acta Crystallog. 5, 802±810. Moreno, E., Teneberg, S., Adar, R., Sharon, N., Karlson, Ê ngstroÈm, J. (1997). Rede®nition of the K. A. & A carbohydrate speci®city of Erythrina corallodendron lectin based on solid-phase assays and molecular modeling of native and recombinant forms obtained by site-directed mutagenesis. Biochemistry, 36, 4429± 4437. Naismith, J. H. & Field, R. A. (1996). Structural basis of trimannoside recognition by concanavalin A. J. Biol. Chem. 271, 972 ± 976. Nicholls, A. & Honig, B. (1991). A rapid ®nite difference algorithm, utilizing successive over-relaxation to solve Poisson-Boltzman equation. J. Comput. Chem. 12, 435± 445. Nicholls, A., Sharp, K. A. & Honig, B. (1991). Protein folding and association: insights from the interfacial and thermodynamic properties of hydrocarbons. Proteins: Struct. Funct. Genet. 11, 281 ± 296. Otwinowski, Z. (1993). In Proceedings of the CCP4 Study Weekend, 29 ± 30 January, 1993 (Sawyer, L., Isaacs, N. & Bailey, S. W., eds), pp. 56 ±62, SERC, Daresbury Laboratory, Daresbury, England. Prasthoffer, T., Phillips, S. R., Suddath, F. L. & Engler, J. A. (1989). Design, expression and crystallization of recombinant lectin from the garden pea (Pisum sativum). J. Biol. Chem. 264, 6793± 6796. Quiocho, F. A. (1989). Protein-carbohydrate interactions: basic molecular features. Pure Appl. Chem. 61, 1293± 1306. Quiocho, F. A., Vyas, N. K. & Spurlino, J. C. (1989). Atomic interactions between proteins and carbohydrates. Trans. Am. Crystallog. Assoc. 25, 23± 35. Ramachandran, G. N. & Sasisekharan, V. (1968). Conformation of polypeptides and proteins. Advan. Protein Chem. 23, 283 ± 438. Reeke, G. N. & Becker, J. W. (1986). Three-dimensional structure of favin: saccharide binding-cyclic permutation in leguminous lectins. Science, 234, 1108± 1111. Reeke, G. N. & Becker, J. W. (1988). Carbohydrate-binding sites in plant lectins. Curr. Topics Microbiol. Immunol. 139, 35± 58. Rini, J. M. (1995). Lectin structure. Annu. Rev. Biophys. Biomol. Struct. 24, 551± 577. Rini, J. M., Hardman, K. D., Einspahr, H., Suddath, F. L. & Carver, J. P. (1993). X-ray crystal structure Ê of a pea lectin-trimannoside complex at 2.6 A resolution. J. Biol. Chem. 268, 10126± 10132.

Rutenber, E. & Robertus, J. D. (1991). Structure of ricin Ê resolution. Proteins: Struct. Funct. B-chain at 2.5 A Genet. 10, 260± 269. SERC Daresbury Laboratory (1994). The CCP4 suite: programs for protein crystallography. Acta Crystallog. sect. D, 50, 760± 763. Shaanan, B., Lis, H. & Sharon, N. (1991). Structure of a legume lectin with an ordered N-linked carbohydrate in complex with lactose. Science, 254, 862 ± 866. Sharma, V. & Surolia, A. (1997). Analyses of carbohydrate recognition by legume lectins: size of the combining site loops and their primary speci®city. J. Mol. Biol. 267, 433± 445. Sharon, N. & Lis, H. (1990). Lectins as cell recognition molecules. Science, 246, 227±234. Sixma, T. K., Pronk, S. E., Kalk, K. H., Wartna, E. S., van Zanten, B. A. M., Berguis, A. M. & Hol, W. G. J. (1992). Lactose binding to heat-labile enterotoxin revealed by X-ray crystallography. Nature, 355, 561 ± 564. Surolia, A., Sharon, N. & Schwarz, F. P. (1996). Thermodynamics of monosaccharides and di-sacccharide binding to Erythrina corallodendron. J. Biol. Chem. 271, 17697± 17703. Toone, E. J. (1994). Structure and energetics of proteincarbohydrate complexes. Curr. Opin. Struct. Biol. 4, 719± 728. Vyas, N. K. (1991). Atomic features of protein-carbohydrate interactions. Curr. Opin. Struct. Biol. 1, 732 ± 740. Vyas, M. N., Vyas, N. K. & Quiocho, F. A. (1994). Crystallographic analysis of the epimeric and anomeric speci®city of the periplasmic transport/ chemosensory protein receptor for D-Glucose and D-Galactose. Biochemistry, 33, 4762± 4768. Wallace, A. C., Laskowski, R. A. & Thornton, J. M. (1995). LIGPLOT: a program to generate schematic diagrams of protein-ligand interactions. Protein Eng. 8, 127± 134. Weis, W. I. & Drickamer, (1996). Structural basis of lectin-carbohydrate recognition. Annu. Rev. Biochem. 65, 441 ± 473. Young, N. M., Watson, D. C., Yaguchi, M., Adar, R., Arango, R., Rodriguez-Arango, E., Sharon, N., Blay, P. K. S. & Thibault, P. (1995). C-terminal post-translational proteolysis of plant lectins and their recombinant forms expressed in Escherichia coli. J. Biol. Chem. 270, 2563± 2570. Zou, J. Y., Flocco, M. M. & Mowbray, S. L. (1993). The Ê resolution of the periplasmic glucose/galac1.7 A tose receptor from Salmonella typhimurium. J. Mol. Biol. 233, 739± 752.

Edited by I. A. Wilson (Received 7 October 1997; received in revised form 19 January 1998; accepted 19 January 1998)