Structural and thermodynamic insights into β-1,2-glucooligosaccharide capture by a solute-binding protein in Listeria innocua

β-1,2-Glucans are bacterial carbohydrates that exist in cyclic or linear forms and play an important role in infections and symbioses involving Gram-negative bacteria. Although several β-1,2-glucan–associated enzymes have been characterized, little is known about how β-1,2-glucan and its shorter oligosaccharides (Sopns) are captured and imported into the bacterial cell. Here, we report the biochemical and structural characteristics of the Sopn-binding protein (SO-BP, Lin1841) associated with the ATP-binding cassette (ABC) transporter from the Gram-positive bacterium Listeria innocua. Calorimetric analysis revealed that SO-BP specifically binds to Sopns with a degree of polymerization of 3 or more, with Kd values in the micromolar range. The crystal structures of SO-BP in an unliganded open form and in closed complexes with tri-, tetra-, and pentaoligosaccharides (Sop3–5) were determined to a maximum resolution of 1.6 Å. The binding site displayed shape complementarity to Sopn, which adopted a zigzag conformation. We noted that water-mediated hydrogen bonds and stacking interactions play a pivotal role in the recognition of Sop3–5 by SO-BP, consistent with its binding thermodynamics. Computational free-energy calculations and a mutational analysis confirmed that interactions with the third glucose moiety of Sopns are significantly responsible for ligand binding. A reduction in unfavorable changes in binding entropy that were in proportion to the lengths of the Sopns was explained by conformational entropy changes. Phylogenetic and sequence analyses indicated that SO-BP ABC transporter homologs, glycoside hydrolases, and other related proteins are co-localized in the genomes of several bacteria. This study may improve our understanding of bacterial β-1,2-glucan metabolism and promote the discovery of unidentified β-1,2-glucan–associated proteins.

␤-1,2-Glucans are bacterial carbohydrates that exist in cyclic or linear forms and play an important role in infections and symbioses involving Gram-negative bacteria. Although several ␤-1,2-glucan-associated enzymes have been characterized, little is known about how ␤-1,2-glucan and its shorter oligosaccharides (Sop n s) are captured and imported into the bacterial cell. Here, we report the biochemical and structural characteristics of the Sop n -binding protein (SO-BP, Lin1841) associated with the ATP-binding cassette (ABC) transporter from the Grampositive bacterium Listeria innocua. Calorimetric analysis revealed that SO-BP specifically binds to Sop n s with a degree of polymerization of 3 or more, with K d values in the micromolar range. The crystal structures of SO-BP in an unliganded open form and in closed complexes with tri-, tetra-, and pentaoligosaccharides (Sop 3-5 ) were determined to a maximum resolution of 1.6 Å. The binding site displayed shape complementarity to Sop n , which adopted a zigzag conformation. We noted that water-mediated hydrogen bonds and stacking interactions play a pivotal role in the recognition of Sop 3-5 by SO-BP, consistent with its binding thermodynamics. Computational free-energy calculations and a mutational analysis confirmed that interactions with the third glucose moiety of Sop n s are significantly responsible for ligand binding. A reduction in unfavorable changes in binding entropy that were in proportion to the lengths of the Sop n s was explained by conformational entropy changes. Phylogenetic and sequence analyses indicated that SO-BP ABC transporter homologs, glycoside hydrolases, and other related proteins are co-localized in the genomes of several bacteria. This study may improve our understanding of bacterial ␤-1,2-glucan metabolism and promote the discovery of unidentified ␤-1,2-glucan-associated proteins.
ATP-binding cassette (ABC) 2 -type transporters are widely distributed in living organisms, forming one of the largest protein superfamilies. ABC transporters utilize the free energy obtained from ATP hydrolysis to import or export a wide variety of molecules across cellular membranes. They share a common architecture consisting of two transmembrane domains (TMDs) and intracellular nucleotide-binding domains (NBDs). In bacterial ABC importers, an additional domain, a solute (or substrate)binding protein (SBP), serves as an initial receptor that specifically binds to ligands with high affinity, delivers them to TMDs, and stimulates the ATPase activity (1). SBPs from Gram-negative bacteria are located in the periplasm, whereas those from Gram-positive bacteria are anchored at the cell surface (2). SBPs are essential for the active transport of their ligands (3,4).
The crystal structures of SBPs have revealed that the domain composition of SBPs is conserved despite their low sequence similarity and widely divergent molecular masses (25-70 kDa) (5). In general, the overall structure comprises two globular ␣/␤ domains consisting of a central ␤-sheet flanked by ␣-helices. The two domains are linked by a hinge, and ligand binding takes place between the two domains. In the absence of ligands, the two domains can move flexibly around the hinge, and an open conformation, in which the two domains are separated, is predominant (6). Upon ligand binding, the two domains get close to each other and are stabilized in a closed conformation. This open-close conformational transition has been called the "Venus Fly Trap" mechanism (7). Although many SBPs have been structurally and func-tionally characterized, full evaluations of many of them remain elusive because of the great diversity of their ligands.
In this study, we focus on the Lin1841 protein to gain insights into the bacterial uptake of Sop n . We show that Lin1841 specifically binds to Sop n s with a DP of 3 or more and clarify its binding thermodynamics and its structural basis by X-ray crystallography, isothermal calorimetry (ITC), and molecular dynamics (MD) simulation.

Binding thermodynamics of Lin1841
The binding thermodynamic parameters of Lin1841 for various carbohydrates (list of ligands under "Experimental procedures") were determined by ITC (Table 1 and Fig. 1). Among the tested oligosaccharides, tight interactions were observed for Sop 3 , Sop 4 , and Sop 5 , and their isotherms exhibited typical sigmoidal curves (Fig. 1, B-D). The binding constants (K a ) were on the order of 10 4 -6 M Ϫ1 , and the binding of Sop 4 showed the highest affinity (K a ϭ 1.72 ϫ 10 6 M Ϫ1 at 20°C), which is in the same range as those of other SBPs (5). No heat pulses were observed for Sop 2 (Fig. 1A), laminaritriose, cellotriose, gentiooligosaccharides, or maltotriose (data not shown), demonstrating that the Lin1841 protein is specific for Sop n s with a DP Ն 3. To date, no SBP that can bind to ␤-1,2-linked glucosides has been reported. The chain length specificity of Lin1841 is consistent with that of LiSOGP (17). Thus, we named Lin1841 the "Sop n -binding protein" (SO-BP). SO-BP also bound to larger linear ␤-1,2-glucans (average DP 25, Fig. 1E). The binding isotherm was not clear, likely due to impurity of the linear ␤-1,2glucan sample. The affinity was estimated to be ϳ10 -1000-fold weaker than that of Sop 3-5 , indicating that SO-BP prefers oligosaccharides rather than polysaccharides. All binding interactions were enthalpy-driven with unfavorable entropy changes, and each of the stoichiometries (n) was almost 1:1. This slight deviation from a 1:1 stoichiometry may be due to impurities of Sop n s. The favorable enthalpy changes increased with an increase in temperature and were not able to sufficiently compensate for increases of the entropic penalties, resulting in slight Binding protein specific for ␤-1,2-glucooligosaccharides decreases of the Gibbs free energy change at higher temperatures (Table 1).
We also measured the change in the molar heat capacity with ligand binding (⌬C p ) by plotting ⌬H versus temperature to estimate the detailed binding thermodynamics of SO-BP (Fig. 1F). The ⌬C p values for Sop [3][4][5] were approximately Ϫ300 to Ϫ200 cal mol Ϫ1 K Ϫ1 . A previous study showed that stacking interactions between one aromatic residue and one sugar ring give ⌬C p values of Ϫ150 to Ϫ100 cal mol Ϫ1 K Ϫ1 (19). Therefore, it was predicted that two or three aromatic residues are involved in the stacking interactions with Sop 3-5 .

Overall structure of SO-BP
The crystal structures of SO-BP in a ligand-free form and in complexes with Sop 3 , Sop 4 , and Sop 5 were determined at 2.2 to 1.6 Å resolutions ( Table 2). The asymmetric units of the crystals contained two molecules. The protein construct used for crys-tallization consisted of residues 26 -422 without the signal peptide. There were no disordered regions in all protein structures, except for the N and C termini; residues 35-420 (ligand-free, chain A), 36 -420 (ligand-free, chain B), 34 - Fig. 2B) structures, attempts to overlay the whole protein in these two states provided C␣ r.m.s.d. values above 1.6 Å. Superimposition of domain II clearly showed a hinge motion (Fig. 2C). A domain movement analysis using the DynDom server (20) indicated that the interdomain rotation angle was 26.1°(97.5% closure), and the translation movement was only 0.3 Å. Although a part of the first ␣-helix (residues 44 -47) of the open state deforms due to the absence of interactions between Glu-45 and the ligands (Fig. 3, B, D, and F), it is likely that the conformational change occurs through rigid body rotation, similar to other SBPs (21). DALI structural homology search revealed that the structure of SO-BP in complex with Sop 4 was most similar to Xac-MalE (a putative maltose/trehalose-binding protein, PDB code 3UOR) from Xanthomonas citri (Xanthomonas axonopodis pv. citri strain 306), with Z-score of 46.6, r.m.s.d. for the 386 C␣ atoms of 3.2 Å, and a sequence identity of 34%. The other DALI hits included a trehalose/maltose-binding protein from Thermococcus litoralis (PDB code 1EU8; Z-score ϭ 43.0, r.m.s.d. for the 373 C␣ atoms ϭ 2.3 Å, and sequence identity ϭ 24%), the solute receptor GacH from Streptomyces glaucescens (PDB code 3K00; Z-score ϭ 42.9, r.m.s.d. for the 370 C␣ atoms ϭ 2.2 Å, and sequence identity ϭ 24%), and the solute-binding protein Lmo0181 from Listeria monocytogenes (PDB code 5F7V; Z-score ϭ 41.8, r.m.s.d. for the 372 C␣ atoms ϭ 2.2 Å, and sequence identity ϭ 20%). When the open form SO-BP was used for the homology search, similar hits were obtained (data not shown).

MD simulation
To explore possible ligand-free conformations of SO-BP that were not observed in the crystal structures, we performed 100-ns MD simulation. Principal component analysis (PCA) was also performed to examine the conformational distribution in the MD trajectory (Fig. 2, E and F). We found two representative structures (state 1 and 2) based on PCA. State 1 (Fig. 2, E and F, shown in orange) was similar to the open state crystal structure (C␣ r.m.s.d. Ͻ0.8 Å), whereas state 2 (shown in blue) was subtly different from the crystal structure (C␣ r.m.s.d. Ͼ1.3 Å). The C␣ r.m.s.d. value and the difference in the interdomain bend angle between states 1 and 2 were 1.1 Å and 8.8°, respectively. Moreover, a domain movement analysis using DynDom indicated that the bend angle difference between the closed state crystal structure and state 2 was 32.5°, suggesting that the ligand-free SO-BP can adopt a more open conformation in aqueous solution.

Architecture of the ligand-binding site
The SO-BP structures obtained from co-crystals with Sop 3-5 s delineated the positions of the corresponding sugars at Binding protein specific for ␤-1,2-glucooligosaccharides the center of the cleft enclosed by domains I and II. We defined glucose units of Sop n as A-E from the nonreducing end ( Fig. 3, A, C and E). The complex structures showed that Sop 3-5 s adopt a zigzag conformation, bind to SO-BP in a similar manner, and share units from the nonreducing end. Units A and B are sequestered from solvent, whereas units at the reducing end (C-E in Sop 3-5 s) are exposed to solvent (Fig. 2B). Two tryptophan residues, Trp-71 and Trp-268, function in stacking with units A and B and the opposite side of unit B, respectively. This observation is in accordance with the ⌬C p values of Sop n s binding, as revealed by ITC (Table 1). Trp-97 forms hydrophobic interactions with the C6 hydroxymethyl group of unit A. Hydrogen bonds play a pivotal role in stabilizing the SO-BP-Sop n complexes (Fig. 3  Binding protein specific for ␤-1,2-glucooligosaccharides hydrogen bonds of Asp-193 and Gln-197 toward unit C appear to be indispensable for ligand binding because Sop 2 was not able to bind to SO-BP (Fig. 1A). The anomeric hydroxyl group of the glucose moiety at unit D in the Sop 4 complex forms a hydrogen bond with the N⑀1 atom of Trp-268. The C4 hydroxyl group at unit E forms an additional hydrogen bond with the side chain of Ser-265. Eighteen-, 36-, and 21-ordered waters are found within 5 Å of Sop 3-5 s, respectively, and thus water-mediated hydrogen bonds also substantially contribute to the recognition of Sop n s. These water molecules are held in place by hydrogen bonds with Met-42, Asp-44, Glu-45, Trp-71, Thr-95, Thr-96, Glu-147, Arg-149, , Met-264, Trp-268, Gly-300, and Glu-377 (Fig. 3, B, D, and F). These features are in accordance with the enthalpy-driven binding modes of Sop n s on SO-BP. In such a manner, the large decrease of enthalpy is attributed to formation of hydrogen bonds, and it compensates for unfavorable entropy changes due to the loss of freedom of water molecules, ligands, and proteins (22).

Contribution of interactions at unit C to complex stability
To estimate the energetic contributions of interactions at unit C to the binding to SO-BP, the difference in free energy changes of binding (⌬⌬G) was calculated based on MD simulations (Fig. 4). We first calculated the ⌬⌬G value between Sop 3 and Sop 2 . This calculation demonstrated that binding of Sop 3 was more energetically favorable than that of Sop 2 , with a ⌬⌬G of Ϫ3.59 Ϯ 0.59 kcal mol Ϫ1 (Fig. 4, A and C). By subtracting the ⌬⌬G value from the binding free energy of Sop 3 at 25°C, the free energy change in Sop 2 binding was estimated to be Ϫ3.16 kcal mol Ϫ1 (K a ϭ 2.08 ϫ 10 2 M Ϫ1 ), which was below the detection limit of ITC. Indeed, no heat signal was observed in the titration of Sop 2 (Fig. 1A). We also calculated the ⌬⌬G value between WT SO-BP and a Q197A mutant. To simply compare the effects of mutation, the ligand-free structures (WT and Q197A) were postulated to be in the closed form. The calculation showed that WT SO-BP exhibited more stable binding Binding protein specific for ␤-1,2-glucooligosaccharides than Q197A, with a ⌬⌬G of Ϫ2.34 Ϯ 0.23 kcal mol Ϫ1 (Fig. 4, B and C). In the same way as described above, the free energy change of Q197A in Sop 3 binding was estimated to be Ϫ4.41 kcal mol Ϫ1 (K a ϭ 1.72 ϫ 10 3 M Ϫ1 ). The calculation for the D193A mutant was not performed because the free energy change for a mutation accompanying a change in the total charge was not calculated accurately.
In addition to the free energy calculations, we determined the energetic contributions of Asp-193 and Gln-197 in Sop 3 binding by ITC (Table 1 and Fig. 4, C-E). A mutation at Asp-193 (D193A) resulted in more than a 4-fold reduction in the K a value (2.01 ϫ 10 4 M Ϫ1 ) compared with WT. A mutation at Gln-197 (Q197A) gave the K a value of 4.34 ϫ 10 3 M Ϫ1 , consistent with the K a value (1.72 ϫ 10 3 M Ϫ1 ) estimated by the above calculation. These computational and experimental results indicate that the interactions at unit C have a significant contribution to the binding free energy of Sop 3 (Ϫ6.75 kcal mol Ϫ1 at 25°C), and the contribution of Gln-197 is larger than that of Asp-193.

Conformation of bound ligands
Distributions in the conformations of Sop n s during MD simulation were analyzed to investigate the difference in the conformations between the Sop n s bound to SO-BP and those free in aqueous solution (Fig. 5). ϭ Ϫ120 to Ϫ80°and ϭ 120 to 180°. The average and angles were Ϫ105 Ϯ 8 and 145 Ϯ 8°, respectively. These results indicate that the interactions between SO-BP and units A and B stabilized the Sop 2 moiety during the MD simulations in a similar conformation as the co-crystal structure (Fig. 5A). In contrast, those of the free Sop 3 s (AB) were distributed in wide ranges, with ϭ Ϫ180 to Ϫ60°and ϭ 60 to 180°. The average and angles were Ϫ91 Ϯ 27 and 120 Ϯ 22°, respectively, which are almost identical to the average values for the free Sop 2 reported previously by simulations and experiments ( ϭ Ϫ83.8 to Ϫ72°and ϭ 110 to 128.5°) (23)(24)(25)(26)(27). Thus, the bound Sop 3 (AB) and the free Sop 3 (AB) were modestly different in their conformations. However, the bound Sop 3 (AB) adopts a sufficiently low free energy conformation according to the free energy map of Sop 2 obtained from the local elevation umbrella sampling (LEUS) method (27). This suggests that the affinity of SO-BP is not significantly influenced by the conformation of Sop 3 . Similar tendencies were observed in the Sop 4, 5 (AB) (Fig.  5, B and C).
The and of the bound Sop 3 (BC) varied but were populated around the Sop 3 (BC) in the co-crystal ( ϭ Ϫ150 to Ϫ90°a nd ϭ 60 to 120°) with the average angles of ϭ Ϫ113 Ϯ 13 and ϭ 84 Ϯ 10° (Fig. 5A). The free Sop 3 (BC) diverged in a similar manner as the free Sop 3 (AB) with the average angles of ϭ Ϫ84 Ϯ 18 and ϭ 97 Ϯ 27°. These angle distributions are also in the stable range shown in the LEUS method. In contrast to the bound Sop 3 (BC), the conformations of the bound Sop 4, 5 (BC) were only populated around the co-crystal structure (Fig. 5, B and C). The and of the free Sop 4, 5 (BC) showed the same conformational properties but did not populate around ϭ Ϫ75 and ϭ 75°like those of the free Sop 3 (BC) (Fig. 5, A-C).
The bound Sop 4, 5 (CD) showed similar conformational properties as the free Sop 4, 5 (CD) and were distributed around the structures of free Sop 2 (Fig. 5, B and C). In contrast to Sop 3-5 (AB, BC, and CD), the conformations of the bound Sop 5 (DE) were extremely divergent and showed a similar conformation to the free Sop 5 (DE). These results suggest that SO-BP restricts the conformations of Sop n s via units A-D site-specific interactions, which makes Sop n s more stable than when free in solution.
Conformations of the hydroxymethyl groups of Sop n s in the co-crystal structures and MD simulations were also examined ( Table 3). Almost all the hydroxymethyl groups of the bound Sop n s (at units A and B both in the co-crystals and simulations) adopted the gg conformation, whereas those of the free Sop n s (A and B) were mainly in the gt conformation. The gg conformation of unit A was fixed by the water-mediated hydrogen bonds and that of unit B was stabilized by the hydrogen bond to the O⑀1 atom of Glu-45 (Fig. 3, B, D, and F). The hydroxymethyl group of the unit C glucose moiety of the bound Sop 3, 4 mainly adopted the gt conformation, which is generally consistent with the free Sop 3, 4 . In contrast, the hydroxymethyl group of unit C of the bound Sop 5 adopted primarily gg (in MD simulation) and gt (in the co-crystal) conformations. During part of the simulation time, the gg conformation was stabilized by a hydrogen bond with the O␦2 atom of Asp-193 (data not shown). The hydroxymethyl groups of units D and E of the bound Sop 4, 5 mostly adopted a gt conformation in the MD simulation. In the co-crystal structures, the hydroxymethyl group of unit D of the Sop 4 adopted a gg conformation and that of unit E of the Sop 5 adopted an intermediate conformation between gt and tg. These hydroxymethyl groups did not form any direct interactions with SO-BP in the crystal structure (Fig. 3, B, D, and F) or during the MD simulations. These differences in the conformation of hydroxymethyl groups appeared to not affect the changes in binding free energy because the free energy difference between gt and gg conformations in glucose is estimated to be 0.1 kcal mol Ϫ1 (28), but it may affect the differences in binding enthalpy and entropy changes for Sop n s.

Distribution of SO-BP homologs and ␤-1,2-glucan utilization loci
We examined the distribution of homologous sequences of SO-BP (amino acid sequence identity Ͼ30%) and found that Binding protein specific for ␤-1,2-glucooligosaccharides orthologs of SO-BP are distributed largely in Firmicutes, Proteobacteria, and Actinobacteria with distinct homology (Fig.  6A). These bacteria are ubiquitously distributed in the environment, such as in soil, food, the sea, and the intestinal tract of animals. SO-BP homologs in Firmicutes are found in various genera, including mainly Listeria, Bacillus, and Paenibacillus. The SO-BP homologs in Actinobacteria are predominantly derived from Streptomyces, and those in Proteobacteria are distributed in almost all ␥-Proteobacteria, including Xanthomonas and its related species.
Furthermore, we examined gene clusters containing putative SO-BP genes to understand the role of SO-BP in ␤-1,2-glucan or Sop n s metabolism. Six representative gene loci were selected  (27), black triangles (26,27), and open squares (23)) or experimental methods (black diamond (24) and open circle (25)). In free Sop 4 (BC), 39% MD snapshots were sampled at ϭ Ϫ100 to Ϫ40°a nd ϭ Ϫ120 to 0°(data not shown). Binding protein specific for ␤-1,2-glucooligosaccharides Binding protein specific for ␤-1,2-glucooligosaccharides to cover all groups in the phylogenetic tree. Two TMDs adjacent to SO-BP are strictly conserved in the gene clusters (Fig.  6B). In the Firmicutes clade, GH94 SOGP and GH3 BGL homologs tends to co-occur in the gene cluster. In the Actinobacteria and Proteobacteria clades, GH1 and GH144 enzymes frequently co-occur in the cluster, and the GH144 enzymes are likely to have endo-␤-1,2-glucanase activity (29). LacI transcriptional regulators also tend to co-localize together with these gene clusters.

Amino acid residue conservation
The degree of amino acid residue conservation of the homologs (Fig. 6A) was mapped on the surface model of SO-BP (Fig.  7A). Highly conserved residues (Fig. 7A, shown in red) are located not only in the binding site but also on the surface of domain I. The conserved patch in domain I corresponds to the region of the maltose/maltodextrin-binding protein from E. coli to interact with the transmembrane MalG (30), implying that these residues may also be responsible for interactions with the associated TMDs of the ABC transporter.
Conservation of the binding site is shown in Fig. 7B. The residues with high conservation scores are mainly located on the side of domain I, and the less-conserved residues are located on the opposite side. Notably, Asp-193 and Gln-197, which are important for ligand binding, are variable in distant homologs (Asp-193 is typically substituted with Pro or Gly), but they are mostly conserved in Listeria and its related species (data not shown).

Discussion
Because ␤-1,2-glucans are not abundant in nature, characterization of ␤-1,2-glucan-associated proteins has not progressed compared with other glucan-associated proteins. Our previous studies overcame this challenge by enzymatically synthesizing linear ␤-1,2-glucan using LiSOGP, leading to identification of ␤-1,2-glucan-degrading enzymes (18,29,31). However, how ␤-1,2-glucan or Sop n s are captured and imported inside the cells remains unknown. In this study, we focused on the Lin1841 protein in the Sop n s utilization locus of L. innocua and revealed its thermodynamic characteristics based on ligand binding and structural analysis. MD simulations supported the structural and thermodynamics data.

Specificity of SO-BP
Our free energy calculations and ITC analysis demonstrate that a key factor determining the specificity for chain length of ligands is the polar interactions toward unit C (Figs. 3, B, D, and  F, and 4). This binding specificity of SO-BP is consistent with that of LiSOGP (GH94 phosphorylase), but the mechanism is different. SO-BP binds to Sop n s with energetically favorable conformations, whereas SOGP from Lachnoclostridium phytofermentans (LiSOGP homolog) is considered to bind to a disaccharide unit of Sop 3 with an unfavorable conformation at subsites Ϫ1 and ϩ1, which is compensated by favorable binding of the third glucose unit at subsite ϩ2 (31).

Binding energetics of SO-BP
Binding of Sop n s are driven by favorable enthalpy changes that are accompanied by unfavorable entropy changes and negative heat capacity changes. Such a calorimetric behavior is generally observed among carbohydrate-interacting proteins (32)(33)(34)(35). This behavior is also supported by the present observations that the Sop 3-5 s are fixed by ordered water molecules and two stacker tryptophan residues (Fig. 3, B, D, and F).
The stacking of the two tryptophan residues is expected to give similar ⌬C p values (Ϫ150 to Ϫ100 cal mol Ϫ1 K Ϫ1 per a sugar-aromatic residue pair) (19). However, the ⌬C p values for Sop 3,4 were approximately Ϫ300 cal mol Ϫ1 K Ϫ1 , whereas that of Sop 5 was approximately Ϫ200 cal mol Ϫ1 K Ϫ1 (Table 1). ⌬C p is a sensitive parameter for changes in the solvent environment, and solvation of polar groups causes negative ⌬C p values (36). In the SO-BP-Sop 5 complex, the reducing end glucose moiety of Sop 5 appears to interfere with the water-mediated hydrogen bond networks around Asp-44 and Asp-193, which is also observed in the Sop 4 complex (Fig. 3, D and F). These polar interactions would contribute to the more negative ⌬C p values for Sop 3, 4 than for Sop 5 .  Fig. 3. Right panel, close-up of the binding site depicts the shape complementarity to Sop n s. B, conserved residues in the ligand-binding site. The amino acid conservation scores were calculated by Consurf using SO-BP homologs (Fig. 6A).

Binding protein specific for ␤-1,2-glucooligosaccharides
The ⌬H and ⌬S values at each temperature increased linearly with DP; therefore, each glucose unit extended from Sop 3 must contribute to the unfavorable ⌬H and the favorable ⌬S (Table  1). Although differences in ⌬S values according to the lengths of Sop n s are dependent on ⌬S conf , a component describing the conformational freedom (Fig. 8), further molecular mechanisms are unclear. It is possible that ⌬H and ⌬S are easily influenced by the effects arising in the bulk solvent, as longer Sop n s would expose their reducing ends more to the outside SO-BP (Fig. 7A). These trends in ⌬H, ⌬S, and ⌬C p were similar to the family 17 carbohydrate-binding module (35). In that study, binding of each glucose unit extended from cellotetraose yielded an unfavorable ⌬H (ϩ1.8 kcal mol Ϫ1 per glucose unit) and a favorable ϪT⌬S (Ϫ2.8 kcal mol Ϫ1 per glucose unit). In addition, the ⌬C p value for cellohexaose was 44 cal mol Ϫ1 K Ϫ1 higher than that for cellopentaose.

Biological implications
Phylogenetic analysis revealed that the ABC uptake system associated with SO-BP is conserved in a number of bacterial species in Firmicutes, Actinobacteria, and Proteobacteria (Fig.  6A). The co-occurrence of GHs and related proteins with the SO-BP homolog and ABC transporter suggest the course of ␤-1,2-glucan dissimilation. A representative degradation system (L. innocua) is schematically shown in Fig. 9. Because SO-BP can discriminate between Sop 3-5 s and ␤-1,2-glucan based on its affinities, this system is highly likely to be dedicated to dissimilation of Sop n s. Other Firmicutes species appear to share this system, as exemplified in Paenibacillus peoriae (Fig.  6B). The co-occurrence of intracellular GH1 enzymes with a SO-BP homolog and an ABC transporter (e.g. Streptomyces and Bifidobacterium in Fig. 6B) is widely found in Actinobacteria species. Considering that the GH1 enzymes mostly act on a ␤-glycosidic bond in exo-mode, these GH1 enzymes adjacent to the SO-BP homolog likely cleave a ␤-1,2-glucosidic bond from the nonreducing end. The absence of an endo-type glycosidase in these loci suggests that each of the loci also targets Sop n s. In the Proteobacteria group (e.g. X. citri and Stenotrophomonas maltophilia), an extracellular GH144 enzyme, which is likely an endo-␤-1,2-glucanase, is present adjacent to a SO-BP ABC transporter homolog (Fig. 6B). In addition, a hypothetical membrane protein (Hypo) co-occurs with the GH144 and SO-BP ABC transporter homolog. This hypothetical membrane protein is predicted to be an outer membrane receptor (TonB_dep_Rec or PF00593 in Pfam), and thus it may assist with translocation of Sop n s across the outer membrane.
In the case of the maltose/maltodextrin utilization gene locus in E. coli, an NBD (MalK) is present with a SBP (MalE) and TMDs (MalF and MalG). MalK is an ATPase responsible for energy coupling to the transport system. However, no NBD gene has been found in the ␤-1,2-glucan utilization loci examined so far. In the L. innocua genome, an NBD gene (Lin0304) is found at a distant locus, and it probably energizes the SO-BP ABC uptake system (Fig. 9).
From these findings, an ABC transporter associated with SO-BP, GH enzyme(s), and a hypothetical membrane protein appear to orchestrate dissimilation of ␤-1,2-glucan or Sop n s. However, we do not know exactly where the above bacteria encounter ␤-1,2-glucan or Sop n s. Several bacteria belonging to the Firmicutes and Actinobacteria groups have genes for an SO-BP homolog and TMDs but not a GH144 gene (Fig. 6B). Therefore, these bacteria may rely on degraded products supplied from other bacteria that have extracellular GH144 enzymes.

Conclusions
This study provides the first structural and biochemical insights into a Sop n s transport protein and will facilitate improved understanding of ␤-1,2-glucan metabolism and the discovery of unidentified ␤-1,2-glucan metabolic proteins. SBPs have the potential to be utilized for biosensor exploitation Binding protein specific for ␤-1,2-glucooligosaccharides (38). Therefore, SO-BP may be applicable as a biosensor for Sop n s (n Ն 3) that are rare in nature.
The lowercase letters in the forward and reverse primers indicate NdeI and XhoI sites, respectively. The genomic DNA was extracted from the cell pellet of L. innocua with InstaGene Matrix (Bio-Rad). The amplified gene was purified, digested by NdeI and XhoI, and inserted into pET30a(ϩ) (Novagen, Madison, WI) to encode a His 6 -tag fusion protein at the C terminus (pET30a(ϩ)-lin1841). E. coli BL21 (DE3) cells (Novagen) were transformed with the constructed plasmid.
The transformant was cultured in Luria-Bertani medium containing 30 mg/liter kanamycin at 37°C until the absorbance at 600 nm reached 0.6. The protein production was induced with 0.1 mM isopropyl ␤-D-1-thiogalactopyranoside at 30°C for 6 h. The transformant cells were collected by centrifugation at 3,900 ϫ g for 10 min and then suspended in 5 ml of 20 mM MOPS-NaOH buffer (pH 7.0) containing 500 mM NaCl (buffer A) per 1 g of the cells. The suspended cells were disrupted by sonication, and the cell debris was removed by centrifugation at 27,000 ϫ g for 30 min to obtain cell extracts. The cell extract was applied to a HisTrap FF crude column (5 ml; GE Healthcare) pre-equilibrated with buffer A. The unabsorbed proteins were removed by washing with buffer A containing 10 mM imidazole, and then the absorbed proteins were eluted with buffer A containing 400 mM imidazole. The eluate was concentrated and buffer-exchanged to 10 mM MOPS-NaOH (pH 7.0) with an Amicon Ultra 10,000 molecular weight cutoff (Millipore). The target protein was applied to a Mono Q 10/100 GL column (GE Healthcare) and eluted with a linear gradient of 0 -500 mM Binding protein specific for ␤-1,2-glucooligosaccharides NaCl in 10 mM MOPS-NaOH (pH 7.0) using an ÄKTA purifier (GE Healthcare). The purity was estimated by SDS-PAGE. The molecular mass estimated by SDS-PAGE (42 kDa) closely corresponded to the theoretical molecular mass (44,531 Da). For ITC experiments, the fractions showing a single band were collected, concentrated, and buffer-exchanged to 20 mM sodium phosphate buffer (pH 7.0). For crystallographic experiments, the fractions showing a single band on SDS-PAGE and a single peak in the chromatogram were collected, concentrated, and buffer-exchanged to 10 mM MOPS-NaOH (pH 7.0). Protein concentration was determined by measurement of the absorbance at 280 nm and calculation from the theoretical extinction coefficients of Lin1841 lacking its signal peptide (72,880 M Ϫ1 cm Ϫ1 ).

ITC
ITC experiments were performed at 15-35°C using Micro-Cal VP-ITC (Malvern Instruments Ltd., Malvern, Worcestershire, UK). The protein dissolved in 20 mM sodium phosphate buffer (pH 7.0) was used for experiments. Ligand solutions were prepared by dilution with the same buffer obtained from the filtrate of the buffer exchange. The protein solution (0.04 -0.3 mM) was stirred at 307 rpm in a 1.44-ml cell and titrated with 5 l of a ligand (0.5-3.0 mM, except for gentio-oligosaccharides and ␤-1,2-glucan; 0.67 and 4.05 mg/ml were used, respectively) 50 times at intervals of 360 or 420 s (except for the Sop 4 titration at 25°C; intervals were set to 600 s). The heat of dilution of the oligosaccharides was determined to be negligible based on control experiments in which the ligand was titrated into buffer solution. Calorimetric data were analyzed using Origin 7.0 software. Thermodynamic parameters, such as the association constant (K a ), the binding enthalpy (⌬H), and the number of binding sites (n), were determined by fitting data into a one-site binding model. It was difficult to determine the accurate ⌬H value toward ␤-1,2-glucan because of uninterpretable heat pulses detected in the ␤-1,2-glucan-titration experiments. The binding Gibbs free energy change (⌬G 0 ), the dissociation constant (K d ), and the binding entropy change (⌬S 0 ) were calculated from the equations ⌬G 0 ϭ ϪRT lnK a ϭ RT lnK d and ⌬G 0 ϭ ⌬H Ϫ T⌬S 0 , where R is the gas constant, and T is the absolute temperature. It is assumed that ⌬H values determined from ITC are equal to the standard enthalpy change (⌬H 0 ). The heat capacity change (⌬C p ϭ ⌬⌬H/⌬T) was calculated from linear regression analysis of ⌬H values at different temperatures. The total entropy change is expressed as the sum of entropy changes in solvent released upon ligand binding (⌬S 0 solv ), conformational freedom around torsion angles of proteins and ligands (⌬S 0 conf ), and the mixing of solute and solvent molecules (⌬S 0 mix ) (⌬S 0 ϭ ⌬S 0 solv ϩ ⌬S 0 conf ϩ ⌬S 0 mix ) (40).

Crystallization and structure determination
All crystals were obtained by the sitting-drop or hangingdrop vapor diffusion method at 25°C. Initial crystallization screening was established using JCSG core suite I-IV and JCSGϩ suite (Qiagen, Hilden, Germany) based on the sitting-drop vapor diffusion method. To obtain crystals in ligand-free form of Lin1841, the crystallization drops were prepared by mixing 0.5 l of 19.3 mg/ml Lin1841 solution with an equal volume of the screening kit solution and equilibrated against 70 l of the same solution. Ligand-free form crystals were generated in a drop of solution containing 0.2 M zinc acetate and 20% (w/v) PEG 3350. After optimizing the conditions, suitable crystals were obtained by mixing 1 l of the protein solution with an equal volume of the reservoir solution containing 0.15 M zinc acetate and 15% (w/v) PEG 3350 and equilibrated against 500 l of the same solution using the hanging-drop vapor diffusion method. These crystals completely grew in 2-3 days.
To obtain crystals of Lin1841 in complex with Sop 3-5 s, the crystallization drops were first prepared by mixing 0.5 l of 27.2 mg/ml Lin1841 solution containing 10 mM Sop 4 in 9 mM MOPS-NaOH (pH 7.0) with an equal volume of the screening kit solution and equilibrated against 70 l of the same solution. Co-crystals with Sop 4 were generated in a drop of solution containing 0.1 M MES-NaOH (pH 6.0) and 40% (v/v) MPD. After optimizing the conditions, suitable crystals were obtained by mixing 1 l of the protein solution with an equal volume of the reservoir solution containing 0.1 M MES-NaOH (pH 5.3) and 42% (v/v) MPD using the sitting-drop vapor diffusion method for 6 days. Co-crystals with Sop 3 were generated in a similar manner as described above, except that Sop 3 was used instead of Sop 4 , and the reservoir solution consisted of 0.15 M MES-NaOH (pH 5.5) and 50% (v/v) MPD. Co-crystals with Sop 5 were obtained using the streak seeding method as follows. A drop was equilibrated for a day in a similar manner as described above, except that 28.1 mg/ml protein containing 5 mM Sop 5 was used, and the reservoir solution consisted of 0.15 M MES-NaOH (pH 5.3) and 42% (v/v) MPD. The co-crystals were generated in the same drop that was streaked with microseeds of the Sop 3 co-crystal for a day.
Crystals in ligand-free form were grown in a drop of reservoir solution supplemented with 30% (v/v) PEG 400 for cryoprotection. There was no need to supply co-crystals with cryoprotectants because of the high concentration of MPD. The crystals were flash-cooled at 100 K in a stream of nitrogen gas. X-ray diffraction data were collected using a charge-coupled device (ADSC Quantum 270) on an NW12A station at the Photon Factory Advanced Ring, High Energy Accelerator Research Organization (KEK), Tsukuba, Japan ( ϭ 1.0000 Å), and processed using HKL2000 (41).
The initial phase of the ligand-free form Lin1841 was determined by the molecular replacement method using MOLREP (42), and the structure of Xac-MalE from X. citri (PDB code 3UOR) was used as a search model. The phase improvement was performed using Morph model on Phenix (43). The initial phase of the Lin1841 in complex with Sop 4 was also determined using MOLREP, and the structure of the ligand-free form was used as a search model. The phase improvement and the automated model building were performed using ARP/wARP (44). Sop 3 and Sop 5 complex structures were solved by the molecular replacement method using MOLREP, and the structure of Sop 4 complex was used as a search model. Initial model structures of ␣-Sop 3 , ␣-Sop 4 , and ␤-Sop 5 were built with JLigand (45). Manual model building was carried out using Coot (46). Crystallo-Binding protein specific for ␤-1,2-glucooligosaccharides graphic refinement was performed using REFMAC5 (47) with the TLS parameters generated by the TLSMD server (48). The refined structures were validated using Molprobity (49) and Rampage (50). The molecular graphic figures were prepared using PyMOL (DeLano Scientific, Palo Alto, CA).

MD simulation
The crystal structures of SO-BP (chain A) in ligand-free form and in complex with Sop 3-5 s were used to construct the initial structures for MD simulations. The N and C termini of proteins were capped with acetyl and N-methyl groups, respectively. The protonation states of histidine were assigned using PROPKA 3.1 (51), and pH 7.0 was used for the calculation. A His-75 of the Sop n complex was protonated on the N␦1 and N⑀2 atoms, and the other histidines were protonated only on the N⑀2 atom. The ligand-free structure and complex structures with Sop 3-5 s were first immersed in cubic water boxes where the distance between protein atoms and the closest boundary was at least 10 Å. Sodium ions were added to the systems for neutralization. The LEaP module of AmberTools 16 (52) was used to produce the initial structures. Amber ff14SB (53) and GLYCAM06j (54) force-field parameters and the TIP3P model (55) were used for the protein, carbohydrate, and water, respectively. The systems were gradually heated to 300 K during 200-ps constant NVT-MD simulations with position restraints on the nonhydrogen atoms of the proteins, and the force constants were set to 10 kcal mol Ϫ1 Å Ϫ2 . During subsequent 800-ps constant NPT-MD simulations, the pressure was adjusted to 1.0 ϫ 10 5 pascal, and the force constants of the position restraints were gradually decreased to 0 kcal mol Ϫ1 Å Ϫ2 . Finally, unrestrained MD simulations were carried out for 100 ns.
The ligand-free SO-BP in the closed form and the free Sop 3-5 s were also equilibrated in aqueous solution. The ligandfree closed SO-BP was prepared by removing the Sop 3 from the SO-BP-Sop 3 complex structure. The structures of the free Sop 3-5 s were prepared using the LEaP module. The simulation procedures were the same as described above, except that the unrestrained MD simulation was performed at 10 ns for the system of the free Sop 3-5 s. The final coordinates were employed as the initial structures of the MD simulations for free energy calculations. All the MD simulations were performed using GROMACS version 5.0.5 (56). In the simulations, the temperature and pressure were controlled by the velocity rescaling method (57) and the weak coupling method (58), respectively. The bond lengths involving the hydrogen atoms were constrained using LINCS algorithm (59), allowing the use of 2-fs time steps. The electrostatic interaction was calculated using the particle mesh Ewald method (60).
PCA was carried out for the MD trajectory of the open form of SO-BP using the method described previously (61). Deviation of the C␣ atoms of the MD snapshots of the trajectory from those of the average structure was analyzed.

Free energy calculation
The contributions of the unit C glucose and Gln-197 of SO-BP to the binding free energy were calculated using alchemical thermodynamic cycles illustrated in Fig. 4, A and B. ⌬G 1 and ⌬G 2 were defined as the difference in the free energy between the free Sop 3 and the free Sop 2 and that between Sop 3 and Sop 2 in the complex, respectively. The contribution of the unit C glucose moiety was calculated as ⌬⌬G ϭ ⌬G 1 Ϫ ⌬G 2 (Fig. 4A). ⌬G 3 and ⌬G 4 were defined as the difference in the free energy between the ligand-free closed forms of SO-BP and its Q197A mutant and between their ligand-bound forms, respectively. The contribution of Gln-197 was calculated as ⌬⌬G ϭ ⌬G 3 Ϫ ⌬G 4 (Fig. 4B). These free energy calculations were based on the Bennett acceptance ratio method (62). In the calculations, a part of the structure was alchemically transformed from one to the other (Sop 3 to Sop 2 or glutamine to alanine) in a stepwise manner, considering 39 intermediate states in each transformation process. In each step of the transformation process, 1-ns MD simulations were performed sequentially, and the last 500 ps of each simulation were used for the free-energy calculation with the Bennett acceptance ratio method. The calculations were performed with the g_bar module of GROMACS.

Other analyses
Sequences for phylogenetic analyses were retrieved from the Refseq database with sequence identities of Ͼ30% (E-value of Ͻ 10 Ϫ60 ) after the BLAST search. The retrieved 57 sequences were aligned using MUSCLE (63), and the phylogenetic tree was constructed using MEGA7 (64) based on the neighborjoining method. Conservation scores of each residues of SO-BP were calculated and colored with Consurf (65, 66) using the above phylogenetic tree and the SO-BP structure in complex with Sop 5 .