Structural Basis Underlying the Binding Preference of Human Galectins-1, -3 and -7 for Galβ1-3/4GlcNAc

Galectins represent β-galactoside-binding proteins and are known to bind Galβ1-3/4GlcNAc disaccharides (abbreviated as LN1 and LN2, respectively). Despite high sequence and structural homology shared by the carbohydrate recognition domain (CRD) of all galectin members, how each galectin displays different sugar-binding specificity still remains ambiguous. Herein we provided the first structural evidence of human galectins-1, 3-CRD and 7 in complex with LN1. Galectins-1 and 3 were shown to have higher affinity for LN2 than for LN1, while galectin-7 displayed the reversed specificity. In comparison with the previous LN2-complexed structures, the results indicated that the average glycosidic torsion angle of galectin-bound LN1 (ψLN1 ≈ 135°) was significantly differed from that of galectin-bound LN2 (ψLN2 ≈ -108°), i.e. the GlcNAc moiety adopted a different orientation to maintain essential interactions. Furthermore, we also identified an Arg-Asp/Glu-Glu-Arg salt-bridge network and the corresponding loop (to position the second Asp/Glu residue) critical for the LN1/2-binding preference.


Introduction
Galectins, β-galactoside-binding proteins, are characteristic of having one or two conserved carbohydrate recognition domains (CRDs) [1,2]. Members of this family have been shown to participate in diverse biological functions, such as cell adhesion, cell growth regulation, and apoptosis via their interactions with β-galactoside-containing structures on cell surface, e.g., N-, O-linked glycoproteins, proteoglycans or glycolipids [3,4]. More importantly, human galectins act as regulatory factors in many types of cancers by either inhibiting or promoting tumor growth [5]. Therefore, to identify selective ligands for human galectins provides not only a useful tool for dissecting how each galectin member interacts with specific glycan structures in correlation with cancer progression, but also a possible solution for the development of clinical therapeutics.
All galectins are able to recognize Galβ1-3/4GlcNAc disaccharides, namely type 1 and 2 Lac-NAc (abbreviated as LN1 and LN2, respectively) that appear in a myriad of glycoconjugates. For instance, LN1 and LN2 are shown as the repeating structures in the non-reducing termini of lacto-series glycans, including blood group antigens. LN2 are constitutively expressed in all mammalian cell types, while LN1 are more tissue-specifically distributed mostly restricted to the epithelia of gastrointestinal and reproductive tract in humans [6]. Interestingly, the presence of LN1 but not the LN2 was recently found to be in association with the pluripotency of induced pluripotent stem cells (iPSCs) [7,8]. Furthermore, studies on the chemical structures of the milk oligosaccharides produced by various mammalian species revealed that LN1-containing oligosaccharides predominate over LN2-containing in human milk [9]. It has been hypothesized to be a selective advantage for human that the acquisition of predominantly LN1-containing oligosaccharides may promote the growth of specific anti-pathogenic bifidobacteria in the infant colon and thus aid their survival [9].
Galectins share major structural homology in their CRDs, but different galectin members were shown to display deviated binding preference for LN1-and LN2-containing glycans [10][11][12][13]. Albeit several crystal structures of galectin/LN2 complexes are available, there is no report regarding to the structure of galectin/LN1 complex. Therefore, the molecular basis still remains enigmatic to underlie the distinct binding preferences of galectins for LN1/2.

Protein preparation
Using a standard PCR-based cloning strategy, the coding region of the full-length hGal1 (residues 1-135), 3 (residues 1-250), 7 (residues 1-136) and CRD domain of hGal3 (residues 113-250) were generated and inserted into modified pET-15b (hGal1, hGal3 and hGal3-CRD) or pET-28a (hGal7) vector (Novagen) with in-frame N-and C-terminal 6xHis-tag, respectively. QuikChange mutagenesis method (Agilent Technologies) was applied to replace the corresponding codons of residues Glu165 and Arg186 in full-length hGal3 with Ala codon to allow the expression of mutant proteins hGal3-E165A and hGal3-R186A. All the wild type and mutated proteins were produced in Escherichia coli BL21 (DE3), with 0.5 mM IPTG induction for 16 h at 20°C. Recombinant proteins were purified by Ni 2+ -affinity and size-exclusion chromatography to homogeneity. Purified proteins were then stored in gel-filtration buffer (25 mM Tris-HCl pH8.0, 300 mM NaCl and 5 mM β-mercaptoethanol) and concentrated to~4, 20 and 9 mg/ml as determined by the method of Bradford for both Biolayer interferometry experiments and crystallization trials, respectively.

Biolayer interferometry
LN1 and LN2 binding affinity of hGal1, hGal3 and hGal7 were quantitatively measured in 96-well microplates at 27°C by Octet Red system (FortéBio). Specifically, biotinylated galectins were prepared according to the standard protocol (provided by FortéBio), and adjusted to final concentration of 1 μM in assay buffer condition (50 mM Tris-HCl pH 7.5, 300 mM NaCl and 5 mM β-mercaptoethanol). Biotinylated galectins were then immobilized on Super Streptavidin Biosensors (FortéBio, Inc.), while free streptavidin sites on biosensor were then blocked by incubation with biocytin (10 mg/ml) to avoid non-specific interactions. The assay was carried out by placing galectin-coated biosensors into the wells with a concentration series of 3-fold diluted LN1/LN2 solutions (200 μl per well, from a top concentration of 3 mM for hGal1 and hGal7, and 1 mM for hGal3) and measuring changes in layer thickness (in nanometers) of biosensors with time. Measurements were followed by 120 sec baseline step, 60 sec association step and 200 sec dissociation step. Baseline and dissociation steps were carried out in assay buffer only. All the data were processed and calculated using Fortebio software and the steadystate K d values were derived from equilibrium responses (S1 Fig) and summarized in Table 1. Two additional parallel Super Streptavidin biosensors were coated with biotinylated galectin and biocytin, separately, and only incubated with assay buffer as double reference controls.

Fluorescence polarization (FP)-based competition binding assay
The measurement of FP-based assay is based on the rotation speed of a fluorophore-containing compound bound to the protein counterpart (e.g. galectins in this study). The fluorophore rotates at a slower rate than when it is unbound, and the resulting fluorescence polarization is higher. In this study we carried out all measurements according to reported procedures [17,18]. To the final sample volume (70 μl) in each assay was added a synthetic FITC-conjugated type 2 LacNAc (LN2-FITC; as the fluorescent probe or reference compound) to a final concentration of 0.1 μM. All the measurements were conducted in Tris buffer (12.5 mM, pH 7.4) with 200 mM NaCl and 5 mM β-mercaptoethanol at 4°C. Data plotting, nonlinear regression analysis, and curve construction was done by Prism 5.0 software (GraphPad, San Diego, CA).
For direct binding assay of LN2-FITC, the data (anisotropy, A vs. hGal 1, 3 and 7 concentration, respectively) were fitted to the formula A ¼ A 0 þ A max Â ½hGal ∕ ðK d þ ½hGalÞ to estimate  The values are determined by Biolayer Interferometry at 300 K. 2 The values are determined by fluorescence anisotropy at 277 K. 3 The binding is too weak to be determined by Biolayer Interferometry. For FP-based competition assay, commercial LN1 (or LN2) molecules (Dextra, UK) at indicated concentration 300, 6 and 300 μM were used as competitors to compete binding interaction between 0.1 μM fluorescent probe (LN2-FITC) and selected concentration of hGal1 (120 μM), 3 (3 μM) and 7 (120 μM), respectively. The measured anisotropy value with competitor (A competitor ) is used to calculate the amount of galectin-bound probe [PG] according to equation: Competitor value was further deduced by the following equations and summarized in Table 1:  and hGal3-CRD-LN1 complexes were obtained by soaking their native protein crystals with 20 mM type I N-acetyllactosamine (LN1) for more than a week. hGal7 protein solution at 9 mg/ml was incubated overnight with 2 mM LN1 molecule at 4°C. Following overnight incubation, the protein-ligand mixtures were centrifuged at 13,000 rpm for 5 min to remove the precipitated protein. Cocrystals for the hGal7-LN1 complexes were carried out by the hangingdrop vapor diffusion method at room temperature by mixing 2 μl protein solution and 2 μl reservoir solution containing 0.1 M Hepes pH 7.5, 0.2 M Li 2 SO 4 , 0.1 M NaOAc and 25% (w/v) PEG 4000. The reservoir solutions supplemented with 10 to 20% glycerol were used for cyroprotection of the complex crystals. The crystals were then flash-frozen in liquid nitrogen and stored for synchrotron-radiation data collection. The diffraction data were processed using the HKL2000 program suite [19] with data statistics as summarized in Table 2.

Determination and refinement of the crystal structures
The crystal structures of all complexes were solved by molecular replacement with the PHENIX AutoMR [20] using previously published ligand-free galectin structures as the starting search models: PDB entries 1W6N [14], 2NMN [21] and 1BKZ [16] for hGal1, 3-CRD and 7, respectively. Model building was performed with PHENIX AutoBuild [20]. The resulting electron density maps were of good quality and show clearly the densities belonging to the bound LN1 molecules. The structure of LN1 molecule was created by JLigand version 1.0.35 in CCP4 software suite [22] and built into the density by using Coot [23]. Structures then underwent rounds of manual model rebuilding and refinement with Coot and PHENIX. Detailed refinement parameters are listed in Table 2. The figures were generated in Pymol [24].

Results and Discussion
Binding affinity and preference of hGal1, hGal3 and hGal7 for Galβ 1-3/ 4GlcNAc We first quantitated the LN1-and LN2-binding affinity of hGal1, 3 and 7 by two different methods, Biolayer interferometry [25] and FP-based competition assays [17,18]. The resulting K d values are summarized in Table 1  whereas hGal1 and hGal3 are more specific for LN2 (Table 1). Biolayer interferometry was known to measure the binding affinity and additional kinetic detail of the given compound/ protein complex, such as kinetic constants of association and dissociation [26]. Recently Biolayer interferometry was applied to characterize the binding properties of galectins [27]. It is common that different methods measuring the same binding event often produce different K d values. The difference could vary from 1 to 2 orders of magnitude [12,[28][29][30], but the most important issue is to see if there is a consistent trend on the binding when comparing diffrerent methods. Overall, the LN1/LN2 binding preference of galectins-1, -3 and -7 are not strict. For instance, only a maximum 3-fold difference in LN1 and LN2 binding affinity of hGal7 was observed in this study, suggesting the possibility of functional redundancy among members of galectin family [31,32]. Given the feature of multivalent interactions in the context of galectin/ glycan lattices, even 2-or 3-fold change in their individual interactions would result in substantially higher activity and/or a more dramatic effect [33]. As a matter of fact, several studies also correlate the binding preference of LN1 or LN2 with physiological activities [34]. Structural information is thus necessary to delineate the insight at molecular level.
Overall structures of hGal1, 3-CRD and 7 in complex with LN1 hGal1, the CRD of hGal3 (hGal3-CRD) and hGal7 in complex with LN1 were crystallized by either a soaking or co-crystallization method. Their crystal structures were then determined between 1.9 and 2.2 Å by molecular replacement on the basis of the published ligand-free structures (PDB IDs: 1W6N [14], 2NMN [21] and 1BKZ [16] corresponding to hGal1, hGal3-CRD and hGal7, respectively) as the starting search models. Statistics of data processing and refinement parameters of the structures are summarized in Table 2. While only one monomer exists in the asymmetric unit of hGal3-CRD co-crystal, each asymmetric unit of hGal1 and hGal7 harbors a distinct symmetric dimer (Fig 1A-1C). Specifically, the dimer of hGal1 exists in a side-by-side manner, whereas hGal7 dimer is present in a back-to-back arrangement. Because the two protomers of the hGal1 and hGal7 dimers are almost identical to each other, we refer to chain A of each crystal structure in the following discussion. All the observed CRDs adopt a typical galectin fold which is composed of two antiparallel β-sheets of six (S1-S6) and five (F1-F5) strands, jointly forming a β-sheet sandwich structure and therefore named as S-sheets and F-sheets, respectively. The S1-S6 β-strands constitute a concave surface to which β-galactoside-containing glycans are bound. Electron density belonging to the bound LN1 is clearly identified in the Fo-Fc electron density map (Fig 1D-1F), indicating that the LN1 molecule in the complex is well ordered and all sugar rings in the LN1 adopt a chair conformation. Of note, overall root mean square deviation (RMSD) values among the newly determined LN1-bound galectins (hGal1, hGal3-CRD and hGal7), the ligand-free and the LN2-bound galectins (PDB IDs: 1W6N, 3ZSM [35] and 1BKZ for ligand-free hGal1, hGal3-CRD and hGal7; PDB IDs: 1W6P [14], 1KJL [15] and 5GAL [16] for LN2-loaded hGal1, hGal3-CRD and hGal7) are quite small. The values of RMSD (C α atoms) between ligand-free and LN1-bound hGal1, 3 and 7 are of 0.57, 0.20 and 0.53 Å, respectively. RMSD values (C α atoms) of 0.10, 0.18 and 0.59 Å are presented between ligand-free and LN2-bound hGal1, hGal3 and hGal7 structures, respectively. Therefore, any differences due to the glycosidic linkages (β1-3 or β1-4) in LN1 and LN2 appear not to seriously distort the overall structure of hGal1, hGal3-CRD and hGal7. Generally the galactose moiety (GAL) forms more H-bonds with the amino acid residues in the CRD than the N-acetylglucosamine moiety (GlcNAc), supporting the idea that GAL serves as a major recognition component. The GAL of LN1 interacts with a series of conserved residues located on S4-S6 β-strands and the loop connecting S4 and S5 strands, which include His44 hGal1 / 158 hGal3 /49 hGal7 , Asn46 hGal1 /160 hGal3 /51 hGal1 , Arg48 hGal1 /162 hGal3 /53 hGal7 , Asn61 hGal1 /174 hGal3 / 62 hGal7 through hydrogen bond (H-bond) networks, and Trp68 hGal1 /181 hGal3 /69 hGal7 via van der Waals contacts (Fig 1G-1I). In particular, the conserved Arg48 hGal1 /162 hGal3 /53 hGal7 residues play an important role in mediating interactions between hGal1, hGal3, hGal7 and their corresponding LN1 molecules, respectively. Specifically, the Arg48 hGal1 /162 hGal3 /53 hGal7 residues not only bridge H-bonds to several oxygen atoms of LN1 including C4-OH, O5 of GAL and C4-OH of GlcNAc, but also connect peripheral carbohydrate-interacting amino acid residues such as Asn46 hGal1 /160 hGal3 /51 hGal7 , Asp54 hGal1 /Glu165 hGal3 /58 hGal7 , and Arg73 hGal1 /186 hGal3 /74 hGal7 to form a characteristic interacting network of H-bonds and electrostatic interactions which are optimal for carbohydrate orientation in the binding curvature (Fig 1G-1I) [16,36]. The density maps of the LN1-interacting amino acid residues are shown in a satisfying quality to define the environment of the LN1-binding site. Structural basis for LN1-and LN2-binding preferences of hGal1, hGal3-CRD and hGal7 In accordance with the X-ray crystal structures, the numbers and distances of specific H-bond interactions involved in the recognition of LN1/ LN2 by hGal1, hGal3 and hGal7 are measured and summarized in Table 3 as the basis to interpret their distinct binding specificity. Generally, hGal1 and hGal3 appear to have shorter distances (in average) in H-bonds to both GAL and GlcNAc moieties of LN2 than those of LN1. In contrast, hGal7 has more H-bonds to GAL moiety and a characteristic shorter distance with GlcNAc in LN1, as compared to those in LN2. These results correlate well with the aforementioned difference in the binding affinity. The static X-ray structures may not always correspond to the behavior of proteins in solution, further studies such as NMR or MD simulations based on these LN1/LN2-galectin complex structures would offer more insights with their binding dynamics in solution.
Overall, the LN1-recognition modes of hGal1, hGal3-CRD and hGal7 are quite similar to those observed in the LN2-bound complex structures. Structural superimpositions of LN2-complexed hGal1, hGal3-CRD and hGal7 structures (PDB IDs: 1W6P, 1KJL and 5GAL, respectively) with their LN1-loaded ones (Fig 2A-2C) indicate that the GAL moiety of LN1 and LN2 is overlapped well. Even though the GlcNAc moiety of LN1 and LN2 interacts with the same amino acid residues, the GlcNAc moiety was found to adopt a different orientation in LN1-and LN2-complex structures. Specifically, the average torsional angles ϕ LN1 (-60°, defined by O5 GAL -C1 GAL -O3 GlcNAc -C3 GlcNAc of LN1) and ϕ LN2 (-66°, defined by O5 GAL -C1 GAL -O4 GlcNAc -C4 GlcNAc of LN2) that depict the position of GAL relative to the glycosidic bond are quite similar (Fig 2D). On the other hand, the torsional angle ψ is to characterize the orientation of GlcNAc relative to the glycosidic bond. The average ψ LN1 (135°, defined by C1 GAL -O3 GlcNAc -C3 GlcNAc -C4 GlcNAc of LN1) is dramatically different from the average ψ LN2 (-108°, defined by C1 GAL -O4 GlcNAc -C4 GlcNAc -C5 GlcNAc of LN2). This~240°shift allows the three important OH groups of galectin-bound LN1 (C4-OH and C6-OH of Gal and C4-OH of GlcNAc) to form the binding interactions not only essential for the gelectin recognition, but also homologous to those produced by the OH groups of galectin-bound LN2 (C4-OH and C6-OH of Gal and C3-OH of GlcNAc) (Fig 2D) [10,37,38]. The glycosidic torsional angles in this study were defined in previous reports [16,36]. Notably the C4-OH group of GlcNAc in LN1 correlates with C3-OH group of GlcNAc in LN2 so that they are located at the equivalent position to form pivotal H-bonds with Arg48 hGal1 /162 hGal3 /53 hGal7 and Glu71 hGal1 /184 hGal3 /72 hGal7 that are highly conserved among human galectins (Fig 2D).
Superimposition of LN1-complexed hGal1, 3 and 7 indicated an important difference in the loop (denoted as L4) between S4 and S5 β-strands (Fig 4), and that Asp54 hGal1 , Glu165 hGal3 and Glu58 hGal7 are differently positioned in L4. L4 of hGal7 is shorter than the counterpart in hGal1 and hGal3 (Fig 4B). The additional amino acid residues in the L4 of hGal1 and 3 thus throng round the space in the vicinity of their carbohydrate-binding sites (Fig 4A). Unlike Asp54 hGal1 and Glu165 hGal3 both situated in the internal of L4, Glu58 hGal7 is resided in either the end of L4 or the beginning of S5 β-strand (Fig 4A and 4B), leading to a differently oriented salt bridge network from those of hGal1 and hGal3 (Fig 2E-2G). Such an arrangement makes it impossible for Glu58 hGal7 to coordinate with the N2 atom of LN2 for additional water-mediated interactions (Fig 3E). Taken together, the LN2-binding preference is possibly linked to the presence of water-mediated interactions, which is under the control of the properly positioned salt-bridge in L4.
Even Asp54 hGal1 and Glu165 hGal3 have a dissimilar location in L4 (Fig 4B), i.e. Asp54 hGal1 is resided at the end of the β-turn structure in L4, while Glu165 hGal3 appears as the first residue of the β-turn. Since Asp54 hGal1 and Glu165 hGal3 residues mediate ionic interactions to their adjacent Arg residues, apparently the location of Asp54 hGal1 and Glu165 hGal3 results in the distinct conformation of L4 in hGal1 and hGal3-CRD (Fig 4A). This explains the reason why hGal3, but not hGal1, binds with tumor-related TF antigen (Galβ1-3GalNAc) [40]. When the x-ray structure of hGal3-TF antigen is superimposed with the hGal3-LN1 structure, the galactose moiety is roughly overlapped (ϕ = -90°vs. -65°, respectively), but the adjacent sugar residue adopts a very different orientation (ψ = 75°vs. 136°). The observation again demonstrates that, in addition to the primary interactions with the galactose residue, the neighboring sugar has to rotate the glycosidic bond to interact well with several important residues, such as Arg162 hGal3 and Glu184 hGal3 . Although Glu165 hGal3 in L4 is not involved in the binding, this residue forms the salt-bridge network to connect with Arg162 hGal3 and Glu184 hGal3 .
Salt bridge motifs are known to play a key role in a number of functions, such as stabilization of protein folds, and arrangement of key residues or waters for the purpose of catalysis or molecular recognition [39,[41][42][43]. In this work the unique salt bridge motifs mediated by Asp54 hGal1 , Glu165 hGal3 and Glu58 hGal7 were identified to coordinate with the vicinal carbohydrate-binding sites to distinguish the bound sugar ligands, which accounts for the LN1/2-binding preferences of hGal1, 3 and 7, respectively.
As discussed previously, the sequence alignment of hGal1, 3 and 7 indicates that L4 is highly variable. But L4 of each galectin appears to be highly conserved among mammalian species (Fig 5), suggesting that L4 is likely to be evolved into different structures. Each of them is well Salt-Bridge Network Affecting Galectin Specificity arranged by heavy H-bond connections between carbonyl and amide groups on their polypeptide backbones, supporting the previous hypothesis that different architectures and dynamics of these variable L4 regions might be functionally relevant for the carbohydrate-binding specificities and thus influence the biological properties of each galectin member [16,40,[44][45][46]. Of particular note, the unique salt bridge networks of hGal1 (Arg48 hGal1 -Asp54 hGal1 -Glu71 hGal1 -Arg73 hGal1 ) and hGal3 (Arg162 hGal3 -Glu165 hGal3 -Glu184 hGal3 -Arg186 hGal3 ) are highly conserved (Fig 5A and 5B), while those in Gal7s are divided into two divergent groups due to the variations in the position of Glu58 hGal7 (Fig 5C). One group of mammalian Gal7s has the corresponding Glu in the end of L4 (hGal7 group), while the other group has this Glu shifted into the middle of L4 (hGal7-like group). Whether the hGal7-like group has a similar LN1-binding preference requires further investigations.
Moreover, it was reported that Gal1 and Gal2 are two most closely related members in the prototype galectin subfamily, and they share up to 43% amino acid sequence identity, in comparison with the 32% sequence identity between hGal1 and hGal7 [44,46,47]. In accordance with the analysis of Hirabayashi et al., both rat galectin-2 and hGal7 displays prominent LN1-preferred binding activity [12]. Multiple sequence alignment of mammalian Gal2s revealed that they are highly conserved ( Fig 5D) and characterized with a shorter L4 sequence that is reminiscent of hGal7-like group (Fig 5E). These analyses strongly suggested that the variable L4 region of galectins is highly relevant to the LN1-or LN2-preferred binding activity. To further demonstrate such relationship, alanine mutation was introduced to replace E165 or R186 of hGal3. Although the binding affinity of hGal3-E165A for LN1 and LN2 was reduced 2.5-and 8.5-fold, respectively (Table 1), the ratios of K d LN1 / K d LN2 were changed from 0.35 of wild-type hGal3 to 1.22 of hGal3-E165A (Table 4), suggesting that the original LN2-preferred binding of hGal3 was shifted to LN1-preference. Obviously the L4-directing Asp/Glu residue is indispensible for affecting LN1/2 binding preference. By mediating distinct salt-bridge network and water-mediated interaction, these charged residues at L4 serve as a key structural element to fine-tune the carbohydrate recognition. Nevertheless, the other site-directed mutant hGal3-R186A was found to lose its binding ability for both LN1 and LN2 although it was prepared with success and was no structurally deviated from the wild type hGal3. The lost binding was realized owing to the close relationship of Arg186 hGal3 with Glu184 hGal3 (a critical LN1/ 2-contact residue) [34,48]. Furthermore, Bonzi et al. identified a conformational change of Arg73 hGal1 (corresponding to Arg186 hGal3 and Arg74 hGal7 ) that was induced by the presence of a unique peptide, λ5-UR of pre-B cell receptor (pre-BCR) [49], leading to the modified To quantitatively evaluate LN1-or LN2-binding preference of hGal1, 3 and 7, LN1/LN2 ratios (K d LN1 / K d LN2 ) were calculated based on the K d values presented in Table 1. 2 The values were obtained based on the K d BI values determined by Biolayer Interferometry. 3 The values were obtained based on the K d FP values determined by FP-based competition assays. 4 It was reported by Hirabayashi, J. et al. [12]. 5 The CRD domain was used instead of full-length hGal3. doi:10.1371/journal.pone.0125946.t004 Salt-Bridge Network Affecting Galectin Specificity carbohydrate-binding specificity of hGal1 (e.g. a 3-fold decrease for Lacto-N-neotetraose (LNnT) and a 10-fold increase for α3-SiaLacNAc). The binding change is crucial for pre-B cell maturation [49,50], suggesting that the distinct salt-bridge network of galectin members are not only relevant to LN1-or LN2-preference, but also critical to determine the binding preference of other ligands.

Conclusions
In summary, our structural studies of hGal1, hGal3-CRD and hGal7 explain how galectins exhibit the binding preference for Galβ1-3/4GlcNAc disaccharides. Since the galactose moiety affords primary interactions, the GlcNAc has to adopt different orientations for keeping comparable H-bonds with several Arg and Glu/Asp residues in a salt-bridge network. Because of the Glu/Asp resided in the variable loop L4, the length of L4 and the location of the Glu/Asp are found to influence the geometry of the salt bridge, resulting in the LN1/ 2-binding preference.
Supporting Information