Legume lectins are essentially metallo-proteins and Erythrina lectins are no exception, they are all Ca²⁺ and Mn²⁺- dependent proteins. Demetallization of the protein has shown that not all tested lectin activities were affected by the presence of chelating agent (50 mM or 100 mM EDTA). This suggests that the metal ions might be buried deep inside the lectin structure and might not affect the binding of the saccharide conjugates. Examples of such lectins are from E. indica, E. caristagalli, E. edulis, E. rubrinervia, and E. americana [10, 12, 21-24] (Table 3). Almost all tested Erythrina lectins were not affected by the presence of EDTA with the exception of E. speciosa and E. senegalensis, where complete inactivation was observed when the protein was treated with chelating agent. The activity of the protein was restored when both Ca²⁺ and Mn²⁺ were added [11, 25]. To emphasize the role of the metal ion in legume lectins, it is found that the binding of the sugar moiety relies on the simultaneous presence of Ca²⁺ and a transition metal ion, Mn²⁺ as in our case. This might be explained by the ability of the calcium ions to stabilize the carbohydrate binding via structural water molecules. The structural analysis of ECorL and ECL showed that each metal ion is attached to four extremely conserved residues and two water molecules. The calcium ion binds Asp129, Phe131, Asn133 and Asp136, while the manganese ion binds Glu127, Asp129, Asp136 and His142. The two aspartate residues form a bridge between the two metal ions [9, 26, 27]. Other residues such as Trp135 are believed to be required for the tight binding of Ca²⁺ and Mn²⁺ to the lectin. This was confirmed by site directed mutagenesis, where mutant-ECorL (W135A) in which Trp135 was replaced by Ala135, resulted in loss of the metal ions upon dialysis and rendered the mutant protein inactive [28]. The structural role of metal ions was studied on the demetalized concanavlin A (Con A) structure, refined at 2.5Å and its subsequent metal ions activation. The study showed that binding of calcium and zinc/cobalt ions is controlled by the key event isomerization of the peptide bond Ala207-Asp208 from trans to cis, and the formation of the transitioned Ca²⁺ binding site by the side chain residues of Glu8, Asp10 and His24. In case of W135A, the release of metal ion occurs due to the increased mobility of the 132-136 loops rather than being coupled directly to the isomerization process observed for Con A. Binding of Zn²⁺ or Co²⁺ in Con A initiates the movement of Asp19 and alters the disordered Pro13-Pro23 loop that result in transforming the lectin from “unlocked” metal free state into “locked” metals bound form, hence enabling carbohydrate recognition [29, 30]. It was also proposed that the presence of high sugar concentration in the absence of metal ions stabilizes the unlock conformation and shift it toward a lock conformation that has higher degree of folding [31], which may be extended to explain in part how binding the monosaccharide can protect against chemotropic exposure. It is noteworthy to say that binding sugar moiety i.e. galactose, to both ECL and ECorL does not bring any conformational change of the D-loop within the binding site of the metalized protein unlike Con A. Hence, the structural effect of the carbohydrate binding to the demetalized ELs remains to be investigated.

Lectins are known for their high stability under harsh conditions. Erythrina lectins retain full activity upon exposure to buffer solutions ranging in pH from 6-10. However, at pH 11, the protein is completely inactive and in acidic buffer solutions, the proteins retain only around 25% activity [11, 14, 20, 25]. On the other hand, full activity was observed in the pH range of 2-10 for E. indica, E. poeppigiana and E. steyermarkii lectins [10, 32]. Konozy et al. investigated the effect of pH on Erythrina lysistemon lectin (ELysL) secondary structure by using Circular Dichroism (CD). It was found that a wide range of conformational changes occurred within the pH range tested. The maximum secondary structure was observed at pH 8.0 and the minimum at pH 2.0 [33]. Thermal studies on some of Erythrina lectins showed that they were stable at wide range of temperatures, up to 70^oC as in the case of ELysL [34]. In general, most of the tested lectins were stable up to 50^oC and rapidly lose their activity to the point of complete abolishment upon reaching the next temperature thermo interval (Table 3). The high stability of lectins might be attributed to the proteins glycosylated, and evidence suggested that glycosylation elevates numerous physical protein instabilities, which occur upon exposure to extreme environmental conditions. Studying the role of glycosylation using Molecular Dynamics (MD) simulations on the structure of E. corallodendron glycosylated-native and N-glycan devoid-recombinant lectins (nECorL and rECorL) showed that glycosylation provides higher thermal stability to nECorL because it affects the side-chain packing of the lectin. This leads to reduction of non-polar solvent accessibility of surface area (ASA) of nECorL compared to rECorL, hence the marginally higher stability [35].

ELs are assembled from one chain protomer of molecular weight ranged between 27 to 35 kDa, which in turn form a dimer of native molecular weight 56-68 kDa. In most cases, the protomers are identical, however hetero dimers are also common (Table 3). The heterogeneity of the subunits’ molecular masses occurs due to differences in the primary structure and/or the variation in the number of the N-glycan chains attached and their structural composition. The chemical structural differences between isolectins were extensively studied on EVLs. Yamaguchi and his colleagues attributed the difference to the substitution at seven amino acids between A-subunit M_r 35 kDa (Gly30, Asn46, Glu71, Tyr106, Val111, Glu202 and Glu206) and B-subunit M_r 33kDa (Asp30, Asp46, Gln71, Ser106, Ile111, Gln202 and Gln206). However, upon treatment with the deglycosylating agent trifluoromethanesulfonic acid, both subunits’ molecular weights were decreased to 31 kDa, which was explained by the loss of the glycan chains [36]. The single polypeptide chain of ELs protomer is composed of 230-294 amino acid residues, which again are within the range of amino acid compositions of legume lectin. Structurally, all ELs are rich in acidic, hydroxyl and aliphatic amino acids residues, with low content of methionine (4-8 residue/Mole lectin) and no cysteine residues, except in the case of E. rubrinervia lectin (ERL), where 8 residues per mole lectin is reported [23] (Table 4). All purified Erythrina lectins reported so far exhibit similar physicochemical properties. They share similarities in the N-terminal amino acid sequences. 16 out of the 18 available sequences of the primary structure of lectin have valine as the first amino acid residue. Only two species display alanine in the N-terminus amino acid. It is to be noted that glutamine and threonine are conserved at position 2 and 3 of the N-terminal sequence (Table 5).

All reported ELs are glycoproteins (natural sugar content 3.6-10%) (Table 3). The N-glycan attached to the lectins is composed mainly of a complex combination of Mannose, Xylose, Fucose and N-acetyl-glucosamine with different ratios and orders. However, the presence of Galactose was reported in E. vespertillo lectin [39] while Arabinose and Galactose were detected in E. indica lectin glycan composition [10] (Table 3). In some studies, Con A was used to determine if the non-reduced terminal of N-glycan is a D-mannose, and was used to purify lectins in some cases [22, 34]. The presence of Xylose is believed to be a signature for plant glycoproteins, as it is believed that the glycans of animal proteins lack the Xylose moiety [40]. Ashford et al stipulated that the N-linked glycan of ECL is characterized firstly by the substitution of the common core of the pentasaccharides [Manα3-[Manα6]-Manβ4NAcGlcβ4NAcGlc] of N-glycoproteins by β-D-Xylose (1→2) attached to the β-mannosyl residue and secondly by the presence of α-L-Fucose (1→3) linked to the reducing end of GlcNAc residue. About 90% of the branched N-glycan was of the following sequence [α-D-Man[1→3]α-D-Man[1→6]]-β-D-Xyl-[1→2]-β-D-Man-β-D-GlcNAc-[1→4]-[α-L-Fuc-[1→3]]-D-GlcNAc [41]. X-ray crystallography studies enabled the generation of an electron density map of the six N-glycan residues out of the seven in the 3D crystal structure of ECL, as the X-ray diffraction resolution was suitable enough to generate such a model [9].

The hepta-N-glycan of ECorL was found to be [α-D-Man(1→3)-[α-D-Man(1→6)]-β-D-Xyl-(1→2)]-β-D-Man(1→4)-β-D-GlcNAc-(1→4)-[α-L-Fuc-(1→3)]-D-GlcNAc], and the heterogeneity of the oligosaccharide moieties occurs due to the lack of Fucose in some of the chains attached to ECorL [42, 43]. The glycosylation site is confined to three amino acid residues, namely Asn17, Asn46 and Asn113. The probable site for attachment of N-glycan to ECL is Asn113, while ECorL and EVA (E. velutina lectin) have two glycosylation sites, the Asn17 and Asn113 in each protomer. However, the crystal structure of ECorL illustrated only the N-glycan joined to Asn17, and the side-chain of this site is structurally compensated by the conformational changes and the formation of hydrogen bonds in rECorL, without affecting the backbone of the site. On the other hand, the three EVL isolectins [with A and B protomers] have different glycosylation patterns and sites. The A-subunit is glycosylated at Asn17, Asn46 and Asn113, whereas the B-subunit is glycosylated at only two sites, Asn17 and Asn113 due to the substitution of Asn47 by Asp47 [9, 27, 36]. Molecular dynamic simulation on ECorL showed that the N-glycan is very mobile, interacts predominantly with the solvent, and it can also form transient non-covalent interaction with the protein surface, at distant amino acids residues (Lys55 and Tyr53), allowing more flexibility of the loops [35, 44-46].

Legume lectins by definition are a family of proteins, which have the same tertiary and importantly variable quaternary structure. X-ray crystallography studies of Erythrina lectins started in 1987 when X-ray diffraction pattern studies were performed on ECorL [46]. This was followed by other successful crystallization and structure determinations of both ECorL and ECL.

ELs, like all other legume lectins are strictly composed of anti-parallel β-sheets, loops, and no α-helices, which form a structure known as “jelly roll”. The β-sheets are divided into three groups, six standard flat back sheets, seven standard curved front sheets and three-to-five small S sheets. In both ECorL and ECL, the first and second back β-sheets are connected and formed via N- and C-terminals. Conversely, the N- and C-terminals of Con A are formed by the front β-sheets. The loops are constructed by around 51% of the total amino acids residues. The main hydrophobic core is located between the front and back sheets and a second hydrophobic nucleus is generated between the front sheets and the extended three loops, forming the metal ion binding site as well as the carbohydrate recognition domain (Fig. 1) [8, 9]. Erythrina lectins possess a unique mode of quaternary association different from the canonical mode of dimerization seen in Con A, or any other legume lectins. They can only exist in the form of dimers linked via handshake association. Such, mode of association was first seen in ECorL and hence called “ECorL-Type”. Many attempts have been made to identify the characteristics of the dimer interface. Elgavish and Shaanan report that the legume lectin dimer is established and stabilized by the chemical composition of the amino acids present in the dimer interface rather than the sequence of the amino acids [47]. Using a graph-spectral algorithm to identify the side-chains and their cluster centers in combination with the traditional multiple sequence alignment indicated that the quaternary mode of association in legume lectins could be determined by their sequence. The ECorL-type of dimerization has an interface called X3 that allows the overlap of the β-sheets of each monomer to form the handshake. Mapping the structurally conserved residues identified in the X3 interface cluster revealed that both SFE and VIKYD motifs located at positions 60-70 and 170 of the amino acid sequence are responsible for the association. ECorL exclusively exists as a dimer and cannot form a tetramer because its sequence lacks the interface determinants at the N- and C-terminals that allow the tetramer formation [48]. It was proposed earlier that the hindrance of the canonical dimer formation in ECorL was due to the glycosylation at Asn17 of the protein N-terminal. However, since other lectins such as ECL, WBA-I and WBA-II exhibit the same handshake mode of dimerization but with a different glycosylation site far from the tetramer interface motifs, it was concluded that the ECorL-type of dimerization occurs due to protein intrinsic factors [9, 44, 49, 50]. Further analysis of the native and recombinant ECorL crystal structures indicated that there is no significant difference except the absence of the N-glycan in rECorL. Both adopt the handshake dimerization, in fact glycosylation only affects the overall thermal stability of the nECorL over the rECorL, not the dimerization properties.

Binding of carbohydrate to lectins is attributed to a shallow groove located on the surface of the molecule known as the carbohydrate binding domain “CBD”. ELs are classified as Hololectins which contain two identical or very similar binding sites [51]. ELs are basically Galactose/Galactose derivative specific lectins (Table 6). The structural features of carbohydrate recognition in ELs were determined from several X-ray crystal structures of both ECorL and ECL, native, recombinant and mutant models. In brief, the CBD is formed by four loops (A, B, C and D) [52]. Loops A, B and C contain the highly conserved residues (Ala88, Asp89, Gly107 and Asn133) which form the binding site of carbohydrate in legume lectin, while loop D is highly variable and its amino acid residues determine the carbohydrate specificity of each lectin. Unlike Man/Glc specific lectin, binding of sugar moieties to ECorL does not induce any conformational change in loop D owing to the presence of water molecules, which would be replaced by a hydroxyl group of the incoming ligand.

Although the binding of lactose to ECL is mediated via different structural arrangements than ECorL [9], the obtained crystals of ECorL in complex with mono and disaccharides reveal that the sugar recognition is taking place in the same manner, with a slight difference presented in additional formation of one direct hydrogen bond and an extra Van der Waals interaction are required when binding Lactose/ N-acetyllactosamine [27]. On the other hand, the structural/thermodynamic relationship studies performed in ECorL binding to disaccharides offer a great number of binding options compared to monosaccharides, which can’t be detected in the X-ray static model [53].

For further studies on the importance of the affinity site amino acids in ECorL, which are likely to affect the affinity and/or the specificity of lectin towards certain sugar moieties, a series of engineered mutant lectins were achieved. The same hemagglutinating activity and affinity towards galactose were obtained after substituting Tyr108 and Pro134-Trp135 residues with equivalent peptide residues found in the peanut agglutinin PNA affinity site (Thr108 and Ser-Glu-Tyr-Asn) and soybean agglutinin SBA (Ser-Trp) [54]. Lower affinity towards GalNAc and almost no affinity for Methyl-β-D-dansylgalactosaminide [MeβGalNDns] are noticed in mutants with the PNA substitution peptide. These results suggest that the differences in sugar specificity between ECorL, SBA and PNA are due to the steric hindrance caused by the additional amino acids found in PNA [55]. According to Adar and her colleagues, Trp135 plays a vital role in enhancement of the affinity binding of lectin to MeαGalNDns [28]. Mutation at position 219 showed that the native Gln219 side-chain provides a conformational flexibility that allows several favorable interactions with LacNAc, hence enhancing the binding affinity [56]. Testing the hemagglutinating activity of the T106G mutant and the subsequent crystallization of lectin with Gal/GalNAc showed remarkable reduction of the lectin affinity towards galactose and, to the contrary, enhancement of GalNAc affinity, these findings are explained by the formation of an additional direct hydrogen bond between GalNAc and the adjacent Gly107 [57].