A structural model for (GlcNAc)2 translocation via a periplasmic chitooligosaccharide-binding protein from marine Vibrio bacteria

VhCBP is a periplasmic chitooligosaccharide-binding protein mainly responsible for translocation of the chitooligosaccharide (GlcNAc)2 across the double membranes of marine bacteria. However, structural and thermodynamic understanding of the sugar-binding/-release processes of VhCBP is relatively less. VhCBP displayed the greatest affinity toward (GlcNAc)2, with lower affinity for longer-chain chitooligosaccharides [(GlcNAc)3–4]. (GlcNAc)4 partially occupied the closed sugar-binding groove, with two reducing-end GlcNAc units extending beyond the sugar-binding groove and barely characterized by weak electron density. Mutation of three conserved residues (Trp363, Asp365, and Trp513) to Ala resulted in drastic decreases in the binding affinity toward the preferred substrate (GlcNAc)2, indicating their significant contributions to sugar binding. The structure of the W513A–(GlcNAc)2 complex in a ‘half-open’ conformation unveiled the intermediary step of the (GlcNAc)2 translocation from the soluble CBP in the periplasm to the inner membrane–transporting components. Isothermal calorimetry data suggested that VhCBP adopts the high-affinity conformation to bind (GlcNAc)2, while its low-affinity conformation facilitated sugar release. Thus, chitooligosaccharide translocation, conferred by periplasmic VhCBP, is a crucial step in the chitin catabolic pathway, allowing Vibrio bacteria to thrive in oceans where chitin is their major source of nutrients.


Edited by Gerald Hart
VhCBP is a periplasmic chitooligosaccharide-binding protein mainly responsible for translocation of the chitooligosaccharide (GlcNAc) 2 across the double membranes of marine bacteria. However, structural and thermodynamic understanding of the sugar-binding/-release processes of VhCBP is relatively less. VhCBP displayed the greatest affinity toward (GlcNAc) 2 , with lower affinity for longer-chain chitooligosaccharides [(GlcNAc) [3][4] ]. (GlcNAc) 4 partially occupied the closed sugar-binding groove, with two reducing-end GlcNAc units extending beyond the sugar-binding groove and barely characterized by weak electron density. Mutation of three conserved residues (Trp 363 , Asp 365 , and Trp 513 ) to Ala resulted in drastic decreases in the binding affinity toward the preferred substrate (GlcNAc) 2 , indicating their significant contributions to sugar binding. The structure of the W513A-(GlcNAc) 2 complex in a 'half-open' conformation unveiled the intermediary step of the (GlcNAc) 2 translocation from the soluble CBP in the periplasm to the inner membrane-transporting components. Isothermal calorimetry data suggested that VhCBP adopts the high-affinity conformation to bind (GlcNAc) 2 , while its low-affinity conformation facilitated sugar release. Thus, chitooligosaccharide translocation, conferred by periplasmic VhCBP, is a crucial step in the chitin catabolic pathway, allowing Vibrio bacteria to thrive in oceans where chitin is their major source of nutrients.
According to the classification proposed by Scheepers et al. (14), SBPs are structurally classified in seven clusters (A, B, C, D, E, F, and G), and cluster C is subdivided into five subclusters (I, II, III, IV, and V) (15). SBPs share a common structural feature comprising two domains connected by a flexible hinge region (16). The flexibility of the hinge enables them to admit specific solutes into the substrate-binding cavity between the two domains. The chitooligosaccharide-specific SBP is localized to the periplasmic compartment of the bacterial cell wall and serves as a mobile component of the (GlcNAc) 2 -specific ABC transporter (5,11). Although the first CBP has been identified in Vibrio cholerae (VcCBP), no functional/mechanistic details of VcCBP were reported (5). Recently, Suginta et al. (13) reported the crystal structure of CBP from Vibrio campbellii (formerly V. harveyi) type strain ATCC BAA-1116 (VhCBP) and the functional characterization indicated that VhCBP strongly binds (GlcNAc) 2,3,4 . VcCBP and VhCBP belong to cluster C/subcluster IV (C-IV), members of which are specific for oligosaccharides, including mannooligosaccharides, cellooligosaccharides, and chitooligosaccharides (5).
Chitooligosaccharide binding to Vibrio CBPs has been proposed to trigger not only translocation of the bound chitooligosaccharides through the inner membrane by the (GlcNAc) 2 -specific ABC transporter but also activation of the gene cluster involved in the chitinolytic cascade, releasing CBPs from chitin-sensor kinase, ChiS, localized to the inner membrane (17). Both biological processes have been shown to require chitooligosaccharide-induced conformational changes by the CBPs. Although the structure and mechanism of a ChiP from Vibrio campbellii (formerly V. harveyi) type strain ATCC BAA-1116 were intensively studied by crystallography in combination with physicochemical and electrochemical techniques (10,12,(18)(19)(20), the molecular mechanism of the periplasmic CBP component involved in sugar translocation and control of gene activation is unknown. In this study, we have focused on the molecular basis of chitooligosaccharide binding and translocation by CBP, a periplasmic SBP specific for chitooligosaccharides. The specific interactions of VhCBP with chitooligosaccharides (GlcNAc) 2-4 were analyzed from the crystal structures. Furthermore, conserved amino acids identified as sugar-binding residues were mutated, and the mutant proteins were characterized with respect to their (GlcNAc) 2 -binding properties. Thermodynamic data obtained from isothermal titration calorimetry (ITC) binding experiments are discussed in the light of crystal structures of the individual VhCBP mutants. Based on these findings, we here propose a 'ping-pong' model for binding/release during (GlcNAc) 2 translocation within the periplasmic space.

Mutational design, recombinant expression, and purification
To identify important amino acid residues in SBPs belonging to the cluster C/subcluster IV, we carried out multiple sequence alignments of five orthologs, including VhCBP, VcCBP, cellooligosaccharide-specific SBP from Thermotoga maritima (TmCeBP), and mannooligosaccharidespecific SBPs from T. maritima (TmMnBP1 and TmMnBP2) (Fig. 1). Based on the crystal structure of VhCBP in complex with (GlcNAc) 2 (13), Glu 10 , Asn 204 , Ser 221 , Trp 363 , Asp 365 , Asn 409, Phe 411 , Arg 436 , Phe 437 , and Trp 513 were found to interact with (GlcNAc) 2 and were therefore selected as targets for mutation. Four of these residues, Asn 204 , Trp 363 , Asp 365 , and Trp 513 , are completely conserved, and the others are partially conserved. VhCBP and the individual mutants were successfully expressed and purified, and all purified VhCBPs migrated as single protein bands on SDS-PAGE. The yields of Structural model of VhCBP in (GlcNAc) 2 translocation the individual proteins from 1 l of the culture medium were 20 to 30 mg for WT and 9 to 15 mg for the mutants. We failed to produce R436A in quantities sufficient for further crystallographic and ITC binding experiments. Arg 436 may be involved in the correct folding of the protein because protein was mostly present in the 'inclusion bodies.' Evaluation of binding affinity of WT VhCBP to (GlcNAc) [2][3][4] We previously reported ITC data for (GlcNAc) [2][3][4] binding to VhCBP in 20 mM potassium phosphate, pH 8.0, at 4 C (13). Here, we re-evaluated the ITC data by conducting binding experiments in Tris HCl, pH 8.0, at 25 C using (GlcNAc) n (n = 2-4) as the ligands. The measurements are summarized in Table 1. The binding interactions for individual ligands were determined assuming a 1:1 stoichiometry. Favorable enthalpy changes (−3.9 to −4.8 kcal mol −1 ) and entropy changes (−6.4 to −4.0 kcal mol −1 ) contributed similarly to the interactions. From these data, we obtained the ΔG values and the dissociation constants (K d ) for (GlcNAc) 2-4 : the binding affinities were fairly dependent on the degree of polymerization of the ligands, with (GlcNAc) 3 exhibiting 3fold and (GlcNAc) 4 12-fold lower affinity than (GlcNAc) 2 .
Crystal structures of VhCBP in complex with (GlcNAc) 3 or (GlcNAc) 4 In the previous article, we reported the mode of binding of (GlcNAc) 2 to VhCBP, based on the crystal structure (13). To further analyze the sugar residue interactions in the groove extending beyond the (GlcNAc) 2 -binding site, we solved the crystal structures of VhCBP in complex with (GlcNAc) 3 and (GlcNAc) 4 . Structural statistics are listed in Table 2. The resolutions of the structures range from 1.5 to 1.8 Å. Figure 2 shows the bound VhCBP structures with (GlcNAc) n (n = 2, 3, and 4) and the apo-VhCBP structure model, obtained by homology modeling using the apo-structure of VcCBP (1ZTY). The overall modeled structure of apo-VhCBP ( Fig. 2A) was different from those of VhCBP with bound (GlcNAc) 2 , (GlcNAc) 3 , or (GlcNAc) 4 (Fig. 2, B-D, respectively), especially in the upper domain colored in blue. The three ligand-bound structures were essentially identical to each other; however, minor conformational differences between the (GlcNAc) 2 -bound and (GlcNAc) 3 /(GlcNAc) 4 -bound structures were found at the GlcNAc-binding site of the lower domain (colored in brown).
Close-up views of the bound (GlcNAc) n are shown in Figure 3. The arrangement of amino acid residues surrounding (GlcNAc) 3 or (GlcNAc) 4 was almost identical to that in the VhCBP-(GlcNAc) 2 complex (13). (GlcNAc) 2 was found at affinity sites 1 and 2 in the protein's binding groove (Fig. 3A), where the nonreducing end of the dimer occupied the innermost affinity site 1, while the reducing end was found at the outer affinity site 2. The electron density of (GlcNAc) 3 (Fig. 3B) in the VhCBP-(GlcNAc) 3 complex was clearly defined at the affinity sites 1, 2, and 3. The main-chain carbonyl of Ala 514 was found to form a hydrogen bond with O6 of the reducing-end GlcNAc of the bound (GlcNAc) 3 . By contrast, the electron density of the reducing-end two GlcNAc residues of the bound (GlcNAc) 4 was weak and uninterpretable (Fig. 3C) at the imaginary sites 3 and 4.
The absence of electron density of pyranose rings may be caused by the disorder of the two GlcNAc residues on the reducing-end side. In fact, the 2Fo-Fc map of the reducing-end GlcNAc of the bound (GlcNAc) 3 was also patchy, allowing only partial occupancy of this sugar ring to be set (Fig. S1). Figure 3D shows the superimposition of the complex structures of VhCBP-(GlcNAc) 2 , VhCBP-(GlcNAc) 3 , and VhCBP-(GlcNAc) 4 . The GlcNAc residues at affinity sites 1 and 2 overlapped completely with each other, indicating identical conformations of the bound two GlcNAc residues at these sites. We noticed that the glycosidic link between the two GlcNAc residues at affinity sites 2 and 3 was twisted, indicating that the link may have a higher conformational energy than that of the neighboring glycosidic link between sites 1 and 2. As noted above, the aromatic side chain of Phe 411 is flipped away from the bound GlcNAc by 3.5 Å in the complexed structures with (GlcNAc) 3 or (GlcNAc) 4 with full occupancy.
Binding mode of (GlcNAc) 4 to VhCBP predicted by docking simulation To predict the structure of the two missing reducing-end residues of bound (GlcNAc) 4 , we conducted docking simulation using AutoDock Vina. Among nine conformers, the most stable provided −10.9 kcal mol −1 of binding energy. However, the dihedral bond angle, defined by atoms O5-C1-O1-C4 at the glycosidic bond between the GlcNAc residues at the affinity sites 3 and 4, was −23.3 . This value is far smaller than those of the most stable conformations reported previously (about −90 ) (21). Therefore, we selected the second-most stable conformer, with −10.5 kcal mol −1 of binding energy, which exhibited reasonable dihedral angles ranging from −96.4 to −73.5 . Figure 4 shows the stereo view of the simulated complex structure.
All thermodynamic values presented in Figure 6A are the mean ± SD, which were obtained from three separate sets of ITC experiments.  Structural model of VhCBP in (GlcNAc) 2 translocation substitute the designated residues and successfully produced nine single mutants, E10A, N204A, S221A, W363A, D365A, N409A, F411A, F437A, and W513A, and one double mutant, W363A/W513A. All proteins were purified to homogeneity by Ni-NTA affinity chromatography, followed by gel filtration, and their purities were checked by SDS-PAGE analysis (Fig. S2). These proteins were used for ITC binding experiments with (GlcNAc) 2 as the ligand. The ITC results are shown in Figure 5, and the thermodynamic parameters are listed in Table 3. The bar diagrams of the thermodynamic parameters, ΔG , ΔH , and −TΔS , are also shown in Figure 6A. All binding interactions showed exothermic profiles (Fig. 5), but the heat generation varied in different mutants. For the double mutant W363A/ W513A, (GlcNAc) 2 titration did not exhibit any clear heat     Figure 6B. Degrees of conservation among five members of cluster/subcluster C-IV SBPs (VhCBP, VcCBP, TmCeBP, TmMnBP1, and TmMnBP2) are also represented by coloring dark red (high degree of conservation) to pale red (moderate degree of conservation) to white (low degree of conservation). As seen from this figure, the effects of mutations on the changes in the binding energy (ΔΔG ) were not always consistent with the degrees of conservation, especially in Asn 204 and Phe 437 .
Crystal structures of the VhCBP mutants in complex with (GlcNAc) 2 To explore the effects of mutations on the structure, the mutated VhCBPs were co-crystallized with (GlcNAc) 2 and single crystals selected for X-ray diffraction experiments. It should be mentioned that the quality of the crystals of E10A, N204A, S221A, D365A, and E363A/W513A in complex with (GlcNAc) 2 was insufficient for X-ray diffraction experiments, and we failed to obtain high-quality datasets for structural determination. Although crystals for all the other mutants were grown under similar conditions, some variations in the crystal packing were observed ( Table 2). Individual structures were solved with the space group of P12 1 1. Close-up views of the (GlcNAc) 2 -binding site of the WT and mutated VhCBPs are shown in Figure 7. Compared with the WT VhCBP (Fig. 7A), W363A not only eliminated the CH-π stacking interaction with the nonreducing end GlcNAc (Fig. 7B) but also slightly shifted the Trp 513 side chain and the bound (GlcNAc) 2 (Fig. S3). These shifts may have altered the interaction of GlcNAc with amino acid side chains from the upper  Figure 5 and are presented as the mean ± SD, which were obtained from three separate sets of experiments. Figure 6. Thermodynamic signature and degrees of conservation of the sugar-binding residues. A, bar diagrams representing ΔG , ΔH , and −TΔS for individual VhCBP mutants. Individual parameter values were obtained from the ITC data shown in Figure 5 and are depicted in blue, green, and red, respectively. B, relationship between reductions in binding affinity by mutations (ΔΔG ) and degree of conservation of individual mutated amino acids. ΔΔG values are stated near the individual mutated amino acids. The mutated amino acids are colored according to the degree of conservation with brightness of red, with color labeled 0.2 for modestly conserved to 1 for totally conserved. The degree of amino acid conservation was calculated by dividing the number of the conserved residues in the sequence alignment by the total number of the corresponding residues in the aligned sequences (Fig. 1). The indicator of the degree of conservation is shown on the upper-right region of this figure. Values presented in Figure 6A are the mean ± SD, which were obtained from three separate sets of ITC experiments. ITC, isothermal titration calorimetry; VhCBP, CBP from Vibrio campbellii type strain ATCC BAA-1116.
Structural model of VhCBP in (GlcNAc) 2 translocation domain, including Trp 513 . The N409A mutation eliminated the hydrogen bond with the reducing-end GlcNAc, but simultaneously, two water molecules fit into the corresponding position (Fig. 7C). One water molecule interacts with O3 of the reducing-end GlcNAc, whereas the other bridges the N-acetyl carbonyl group and Nη of Arg 436 . The F411A mutation did not affect the hydrogen bonds but eliminated the hydrophobic interaction with the N-acetyl methyl group of the reducingend GlcNAc (Fig. 7D). A similar effect was observed in F437A, that is, the hydrophobic interaction with the N-acetyl methyl group of the reducing-end GlcNAc was likely eliminated by the F437A mutation (Fig. 7E); however, in this case, the hydrogen bonding network formed by the Arg 436 side chain was partly affected by the Ala mutation of Phe 437 . As seen in Figure 7F, the W513A mutant was found to eliminate the hydrogen bond with the acetamido -NH 2 of the nonreducing end GlcNAc and the CH-π stacking interaction with the pyranose ring of the reducing-end GlcNAc, most likely causing the strong reduction of the binding affinity (ΔΔG , +4.2 ± 0.3 kcal mol −1 ). A remarkable structure was seen with W513A, in which the substrate-binding groove was partially open at the reducing-end side (site 2). The distance between C α carbons of Ala 24

Discussion
The mode of binding of (GlcNAc) 3

and (GlcNAc) 4 to VhCBP
In most carbohydrate-related proteins that have long and extended binding grooves, the binding affinities depend strongly on the degree of polymerization of the oligosaccharide substrates (22); the higher the polymerization degree of the oligosaccharide, the greater the binding affinity. However, we report here that VhCBP binds (GlcNAc) 2 , (GlcNAc) 3 , and (GlcNAc) 4 with ΔG values of −10.3 ± 0.3, −9.7 ± 0.3, and −8.8 ± 0.3 kcal mol −1 , respectively ( Table 1). The binding affinity did not correspond to the degree of polymerization and was the lowest in the longest ligand, (GlcNAc) 4 . To rationalize the unique dependency on the degree of polymerization of ligands, we analyzed the crystal structures of VhCBP in complex with (GlcNAc) 3 and (GlcNAc) 4 (Figs. 2 and 3). A twist of the glycosidic linkage was observed between the GlcNAc residues at the affinity sites 2 and 3 of the bound (GlcNAc) 3 (Fig. 3B), suggesting a potential loss in the stabilization energy of the complexed structure. Furthermore, the Phe 411 side chain was found to flip away from the bound GlcNAc (3.5 Å) in the complex with (GlcNAc) 3 (Fig. 3D). A water molecule interacting with the carbonyl group of Ala 514 through a hydrogen bond is replaced by O6 of the reducing-end GlcNAc (affinity site 3) in the VhCBP-(GlcNAc) 3 complex. This can be seen from the modeled structure of the VhCBP-(GlcNAc) 4 complex (Fig. 4). No other interactions were found in the third GlcNAc, indicating only a weak interaction of the GlcNAc at site 3. This was confirmed from the patchy electron density of the reducing-end GlcNAc (Fig. S1), suggesting flexibility of this GlcNAc. The enthalpy/entropy gain derived from the additional hydrogen bond and the flexible GlcNAc residue at site 3 appear to be almost compensated by the disadvantages incurred by the flipping of the Phe 411 side chain and twisting of the glycosidic bond, resulting in affinity for (GlcNAc) 3 that is similar to that for (GlcNAc) 2 .
The electron density of (GlcNAc) 4 on the reducing-end side was too weak to place the two GlcNAc moieties at the imaginary affinity sites 3 and 4 (Fig. 3C). This suggests that the mobility of the fourth sugar has a knock-on effect to weaken the binding of the third sugar. Flipping of the Phe 411 side chain was also observed in the complex with (GlcNAc) 4 , as in the complex with (GlcNAc) 3 , and appears to widen the space at sites 3 and 4. It is most likely that GlcNAc residues are present at sites 3 and 4 and are disordered because of the large space on these two sites. Docking simulation of the structure of VhCBP in complex with (GlcNAc) 4 revealed that interactions are formed mostly with the two GlcNAc units at the affinity sites 1 and 2, whereas on sites 3 and 4, only two hydrogen bonds are formed, by the main-chain carbonyl of Ala 514 and the Oδ side chain of Gln 102 , respectively (Fig. 4). The middle glycosidic linkage of the bound (GlcNAc) 4 was also twisted, as in (GlcNAc) 3 bound to VhCBP (Fig. 3B). The disadvantage derived from the twist of the glycosidic linkage appears to predominate over the advantages derived from additional interactions/disorder of the GlcNAc residues at sites 3 and 4, resulting in the lower affinity for (GlcNAc) 4 . A similar observation was reported for TmCeBP in complex with cellopentaose, where cellopentaose was bound through water molecules, but not by direct interaction at the same affinity sites (23). A longer and more versatile binding cleft was observed for maltose-binding protein (MBP) from Escherichia coli (24,25). MBP was identified to be able to bind maltooligomers of 2 to 7 glucose units. The crystal structure of MBP in complex with maltodextrin indicated that the strong affinity sites of MBP are more extended to sites S1, S2, and S3, enabling maltotriose to be tightly accommodated (25). Asp 365 , Trp 363 , and Trp 513 are completely conserved among the cluster C-VI SBPs (Figs. 1 and 6B), suggesting their importance in the binding of sugars to SBPs. We already reported the importance of the two tryptophan residues 16 , which appear to sandwich the oligosaccharide ligand (Fig. 6B). The mutational effects of Trp 363 and Trp 513 on ΔG were similar (ΔΔG = +4.7 ± 0.4 and +4.2 ± 0.3 kcal mol −1 , respectively), but the effects of mutations on enthalpy/entropy contributions are significantly different between the two Trp mutants. The W363A mutation had a large effect on the enthalpic term, while in W513A, a large change was observed in the entropic contribution (Table 3). This can be attributed to the difference in the interaction modes between the two residues. Trp 363 interacts purely through CH-π stacking, while Trp 513 interacts through both CH-π stacking and hydrogen bonding with the N-acetyl group of the GlcNAc in site 1. As demonstrated from the crystal structure of W363A (Fig. S3), the Trp 513 side chain is shifted by the mutation of Trp 363 by 0.6 Å, and the bound (GlcNAc) 2 also shifted slightly. These rearrangements probably affect the binding interactions provided by amino acids in the upper domain, including Glu 10 , Asn 204 , Ser 221 , and Trp 513 . Similar phenomena may take place on mutation of Trp 513 . These effects may enhance ΔΔG caused by the mutations of the two important tryptophan residues.
The greatest impact on ΔG was observed in the Asp 365 mutation (ΔΔG = +8.6 ± 0.3 kcal mol −1 ). Specifically, this mutation affected the entropic term much more than the enthalpic term, not only eliminating the two hydrogen bonds that were observed in the crystal structure (13) but also affecting the state of solvation. The number of bound water molecules around Asp 365 was strongly reduced by the mutation to alanine; hence, the entropy gain due to the desolvation upon GlcNAc residue binding was much lower in the D365A mutant than in the WT. This enhanced ΔΔG in the D365A mutant. By contrast, the reduction in hydrophobicity in the F437A mutation may not only have weakened hydrophobic interaction with the N-acetyl methyl group of the reducing-end GlcNAc at the site 2 but also added a conformational restraint to Phe 411 . The F437A mutation may have also affected the hydrogen bond network formed by the nearby amino acid, Arg 436 , with the N-acetyl carbonyl group of the same GlcNAc and with a bound water molecule. These subtle changes resulted in a rather large impact on ΔG (ΔΔG = +4.1 ± 0.3 kcal mol −1 ). It is notable that Phe 437 is conserved only among the chitin-specific SBPs, that is, VcCBP and VhCBP, and Arg 436 and Phe 437 recognize the N-acetyl group of the ligand. The specificity toward GlcNAc may be mainly attributed to these two amino acids, indicating that mutational study of Arg 436 and Phe 437 may lead to specificity-engineered sugar-specific SBPs. Here, we concluded that the major contributors, Asp 365 , Phe 437 , Trp 363 , and Trp 513 , do not simply interact with GlcNAc but control the state of the surrounding amino acids or bound water molecules.
Minor contributions of Glu 10 , Asn 204 , Ser 221 , Asn 409 , and Phe 411 The other mutational effects (of E10A, N204A, S221A, N409A, and F411A) on ΔG were moderate, although one mutation involved the amino acid with the highest degree of conservation, Asn 204 (Fig. 6B). Glu 10 , Asn 204 , and Ser 221 interact with the GlcNAc residue at site 1 with a complicated hydrogen bonding network in the narrow binding cavity. The effect of eliminating the interactions of one of these amino acids may be compensated by the interactions of the other amino acids located close to the bound (GlcNAc) 2 ; for example, the interaction of O4 of the nonreducing GlcNAc is shared by Asn 204 and Asp 365 , while O3 of the same GlcNAc is shared by Asn 204 and the main-chain NH of Phe 222 . Thus, the mutational effect of Asn 204 is compensated by Asp 365 and Phe 222 . Similar observations were made for Glu 10 and Ser 221 . A hydrogen bond of Asn 409 with O3 of the GlcNAc at site 2 is eliminated by the N409A mutation; however, two water molecules were introduced into the VhCBP-(GlcNAc) 2 interface. This may be the cause of the decreased enthalpic term and increased entropic term listed in Table 3. Finally, the N409A mutation resulted in a small decrease in the binding affinity. Mutation of Phe 411 lowered the favorable contribution of the enthalpy change, suggesting a role of Phe 411 in masking the Nacetyl group from water molecules, avoiding direct contact with water and forming a hydrophobic interaction with the Nacetyl methyl group.
The chitin catabolic cascade of Vibrio spp. was previously proposed by Li and Roseman (5) to be rigorously controlled by the environment signal (GlcNAc) 2 . Figure 9A illustrates how (GlcNAc) 2 plays a central role in the rigorous control of the chitin catabolic cascade. (GlcNAc) 2 and other chitooligosaccharides are transiently generated from chitin degradation by chitinase (26,27) and then transported across the outer membrane through ChiP. The final degradation products (GlcNAc) 2 and GlcNAc are generated by periplasmic enzymes, chitin dextrinase and β-GlcNAcase, respectively (5). Binding of (GlcNAc) 2 releases the CBP from a membrane-bound histidine kinase sensor (ChiS), which essentially activates the gene cluster that is involved in the downstream cascade. CBP translocates (GlcNAc) 2 to the (GlcNAc) 2 -specific ABC transporter, for further energy production. An alternative route of chitin catabolism involves GlcNAc transport through a general diffusion porin, followed by (GlcNAc)-specific phosphoenolpyruvate:carbohydrate phosphotransferase system permease. In the chitin catabolic cascade, (GlcNAc) 2 acts as the intrinsic signaling molecule and a key controller of the chitin catabolic cascade.
Although we failed to obtain the crystal structure of apo-VhCBP, the crystal structure of the apo-VcCBP from V. cholerae (1ZTY), which is highly homologous to VhCBP (83% identity) was observed in a 'fully open' form (Fig. 8A), in which the upper and lower lobes are widened. Binding of (GlcNAc) 2 induced domain movement, as observed in the crystal structure of VhCBP in complex with (GlcNAc) 2 (5YQW) 16 and agreeing with the structure of VcCBP in complex with the same substrate (1ZU0). Such a 'sugarinduced-fit' produced a 'fully closed' state. The most interesting finding in our study was made with W513A in complex with (GlcNAc) 2 . Trp 513 is located at the mouth of the GlcNAcbinding groove, and its position is thought to regulate the translocation of (GlcNAc) 2 to the (GlcNAc) 2 -specific ABC transporter. In fact, the crystal structure of the W513A mutant showed a 'half-open' conformation (Fig. 8C), which effectively provides a snapshot of the sugar-releasing process that facilitates translocation. An uptake system for an acidic polysaccharide, alginate, in Sphingomonas sp. A1 was thoroughly investigated by crystallography (28,29). The crystal structure of alginate-specific SBP (AlgQ2) in complex with an alginatespecific ABC transporter (AlgM1M2SS) was solved in the presence of alginate oligosaccharide. AlgQ2 and AlgM1M2SS form a tunnel-like binding cavity at their interface, and the nonreducing end of bound alginate oligosaccharide interacts strongly with the inner part (closed end, affinity sites 1 and 2) of the binding cavity in the closed conformation of AlgQ2. However, the reducing-end residue is exposed from the exit of the tunnel-like binding cavity (open end, affinity sites 3, 4, 5…) and the oligosaccharide appears to be translocated into the charged surface of the transmembrane domain of AlgM1M2SS (29). Although the sequence similarity is quite low, the 'halfopen' conformation of W513A obtained for VhCBP is analogous to AlgQ2 in complex with AlgM1M2SS. VhCBP interacts strongly with (GlcNAc) n at the nonreducing end, while the reducing end interacts loosely or unfavorably with VhCBP. After binding to the corresponding transporter, VhCBP in complex with (GlcNAc) n may undergo a subtle change in its conformation from 'closed' to 'half-open' by eliminating the interaction with Trp 513 . This half-opening may further enhance the looseness of the interaction, facilitating the translocation of (GlcNAc) n . The bound (GlcNAc) n may then be released from the substrate-binding groove, like a ball bouncing off a racket. Similar observations were reported for other oligosaccharide-specific SBPs, such as TmCeBP (23) and E. coli MBP (24,25). This common feature of binding grooves in SBPs suggests an energetic advantage of the 'half-open' binding groove on the reducing-end side on the sliding motion of oligosaccharides during sugar translocation.
The structural data obtained from our previous report 16 and current studies together suggest the 'ping-pong' model for (GlcNAc) 2 binding/release as depicted in Figure 9B. Such a Structural model of VhCBP in (GlcNAc) 2 translocation model involves three steps: step 1, the apo-VhCBP in the 'fully open' form binds (GlcNAc) 2 , which enters the substratebinding groove, its binding inducing a 'fully closed' conformation of VhCBP.
Step 2, the fully closed conformation subsequently undergoes transition to the intermediate 'halfopen' conformation, and in step 3, (GlcNAc) 2 leaves the binding sites, triggering the return of VhCBP to its original open conformation. The crystal structure of the W513A-(GlcNAc) 2 complex observed in the present report is the first experimental evidence capturing an intermediary conformation between fully closed and fully open conformations in the sugar translocation pathway. Our ITC data suggested the closed (sugar-bound) state has high affinity to the (GlcNAc) 2 substrate (K d of 0.03 μM), while the semiopen state has 1200 lower affinity than the closed state (K d of~37 μM). We propose that this decrease in the binding affinity facilitates the sugar release from the VhCBP to its membrane-transporting partner. In conclusion, the proposed structural model is thought to be a general mechanism in the Vibrio system, facilitating the translocation of small chitooligosaccharides from the periplasm to inner membrane-transporting components, with (GlcNAc) 2 being the preferred molecule, allowing the bacteria to utilize chitin as their source of energy.

Multiple sequence alignment of SBPs
VhCBP was found to belong to cluster C/subcluster IV (C-IV) (15). Four other SBP members occur in the same cluster/ subcluster, including VcCBP from Vibrio cholera (PDB code, 4GFR), a TmCeBP (PDB code, 2O7I), and TmMnBP1 and TmMnBP2 (PDB codes, 4PFY and 4PFW, respectively). Amino acid sequences of VhCBP (PDB code 5YQW) and the four well-defined sugar-specific SBPs mentioned above were aligned by MUSCLE (29,30). The secondary structural elements of VhCBP were integrated and displayed on top of the aligned sequences using the program ESPript v. 3.0 (31) (Fig. 1).

Construction of the mutated proteins of VhCBP
Mutant plasmids were constructed using the previously generated pET23d(+)-cbp from V. harveyi as a DNA template (13). Mutation was introduced into the WT gene using a sitedirected mutagenesis kit (Stratagene). Primers used for sitedirected mutagenesis are listed in Table S1. Mutation sites are shown in bold, and genetic codons corresponding to the mutated residues are indicated by single underlining. For the double mutant W363A/W513A, the W363A mutation was introduced using the plasmid encoding W513A mutant as a DNA template. Correct mutations were verified by DNA sequencing service (Macrogen, Inc).

Expression and purification of WT VhCBP and mutants
The competent cells of E. coli Origami (DE3) (Novagen) were transformed with the recombinant plasmids harboring the WT or mutated VhCBP genes. Each transformant was inoculated into the LB medium containing 100 μg ml −1 ampicillin and shaken at 37 C until OD 600 reached 0.8. Protein expression was induced by the addition of 0.05 mM IPTG to the cell culture that had been precooled on ice for 10 min. The IPTG-induced cell culture was shaken at 18 C for a further 24 h, after which the cell pellet was harvested by centrifugation at 4500 rpm, 4 C for 45 min, and the separated culture medium was discarded. The cell pellet was resuspended in 20 mM sodium phosphate buffer, pH 7.4, containing 150 mM NaCl, 10% (v/v) glycerol, 1 mM PMSF, and 0.1% Triton X-100, and the cells in the suspension were disrupted with an ultrasonic processor (Cole-Parmer Instrument Company, LLC). The cell lysate was harvested by centrifugation at 12,000 rpm at 4 C for 45 min. For protein purification, the crude cell lysate was applied to a Ni-NTA column equilibrated with 20 mM sodium phosphate, pH 7.4, containing 150 mM NaCl, then washed twice with the buffer containing 10 mM imidazole, and eluted with the buffer containing 100 mM imidazole. Eluted protein fractions were collected and diluted three-fold with the same buffer and applied to a HiTrap Q FF column 5 ml (GE Healthcare). The column was washed twice with 20 mM sodium phosphate buffer, pH 7.4, containing 50 mM NaCl and adsorbed protein then eluted with a linear gradient of 50 to 500 mM NaCl. Protein fractions containing VhCBP or the mutants were collected, concentrated by Amicon Ultra-15 (Millipore), and applied to a HiPrep 16/60 Sephacryl S-100 column (GE Healthcare) equilibrated with 20 mM sodium phosphate buffer, pH 7.4, containing 150 mM NaCl. Proteins were isocratically eluted and fractions collected and analyzed by SDS-PAGE. Fractions containing VhCBP or the mutant were pooled and stored at 4 C for further analysis. Protein concentration was determined from absorbance at 280 nm, using a molar extinction coefficient calculated from the equation proposed by Pace et al. (32).  Table S2. Briefly, the temperature in the sample cell was maintained at 25 C, and the difference in temperature (ΔT) between the reference cell and the sample cell was compensated by heaters fitted to the cells. The compensation energy utilized per second was plotted against experimental time. Peak integrations of the individual troughs were replotted against the molar ratio of the ligand to protein. The binding isotherm thus obtained was used for data-fitting analysis using the 'One-Set of Sites' model defined in MicroCal PEAQ-ITC Analysis Software to obtain K d , ΔG , ΔH , and ΔS terms. Stoichiometry n was fixed at 1.0 because the crystal structure of VhCBP in complex with (GlcNAc) n clearly indicated a 1:1 (protein:ligand) interaction. Titrations of (GlcNAc) 2 to the individual VhCBP mutants were performed as described for the WT protein. Because affinity and heat generation changed depending on the protein used, ligand and protein concentrations used for the experiments were varied (Table S2) to fulfill the criteria proposed by Turnbull and Daranas (33).

ITC binding experiments
Crystallization and structure determination of VhCBP mutants in complex with (GlcNAc) 2 Crystallization conditions for individual mutants were screened using Morpheus (Molecular Dimensions). Protein solution (10 mg ml −1 ) was freshly prepared in 10 mM Hepes buffer, pH 7.0, containing 100 mM NaCl and mixed with an equal volume of the precipitant solution to make a crystallization drop. Crystal screens were set up by a microbatchunder-oil technique using 100% (v/v) silicone oil (Hampton Research) at 20 C. Single crystals were harvested from crystallization drops by a cryoloop and flash-frozen and stored under liquid nitrogen. VhCBP was successfully cocrystallized with (GlcNAc) 3 and (GlcNAc) 4 . Although cocrystallization with (GlcNAc) 2 was attempted with all the mutants produced, only five mutants, W363A, N409A, F411A, F437A, and W513A, were cocrystallized with (GlcNAc) 2 under the conditions described in Table S3. With the other mutants, crystallization was unsuccessful.
X-ray diffraction experiments were performed at the beamline, TPS-05A, in NSRRC (National Synchrotron Radiation Research Center, Taiwan). An automated system controlled by the software installed in the Linux machines at the beamline handled the crystals during experiments. Briefly, crystals were mounted onto the goniometer automatically by a robotic arm and irradiated with X-ray at λ = 0.999840 Å, and the diffraction images were collected by MX300-HS (Rayonix, LLC), integral to the beamline. Diffraction images were analyzed by HKL2000 (34). The resulting intensity file was transformed to structural factor file and phased by molecular replacement using the complex structure of VhCBP and (GlcNAc) 2 (PDB code, 5YQW) as the template by PHASER MR in CCP4i program suite v. 7.0 (35,36). The modeled structures were refined by a combination of Coot (37) and Phenix.Refine (38)(39)(40)(41) in Phenix 1.9 (42). The statistics of data collection and structural refinement are shown in Table 2. The structures were displayed by PyMOL Molecular Graphics System v. 2.4.0. Structures were successfully solved with the highest resolutions below 2.2 Å and deposited in the PDB database under the PDB codes 6LZQ for VhCBP-(GlcNAc) 3 , 7EBI for VhCBP-(GlcNAc) 4 , 7EBM for W363A-(GlcNAc) 2 , 6LZT for N409A-(GlcNAc) 2 , 6LZU for F411A-(GlcNAc) 2 , 6LZV for F437A-(GlcNAc) 2 , and 6LZW for W513A-(GlcNAc) 2 .

into VhCBP
Because the electron density of two sugar residues on the reducing-end side of bound (GlcNAc) 4 was not observed, (GlcNAc) 4 was modeled into the VhCBP structure extracted from the VhCBP-(GlcNAc) 4 complex (7EBI), generating the most plausible structure of the VhCBP-(GlcNAc) 4 complex. AutoDock Vina was used to simulate the complex structure (43,44). The ligand was prepared using GLYCAM and PRODRG (45,46). The ligand was docked into the VhCBP molecule using Iterated Local Search global optimizer in AutoDock Vina. Conformers were searched within a grid box with 40, 40, and 30 Å for each dimension covering the entire binding pocket of VhCBP. Finally, the conformation was converged to nine conformers with the energies calculated, −10.9, −10.5, −10.1, −9.3, −9.0, −8.8, −8.6, −8.5, and −8.5 kcal mol −1 , respectively, from the most stable to unstable conformer.
Supporting information-This article contains supporting information with supplementary experiments (Figs. S1, and S2) and full wwPDB X-ray structure validation reports for all the crystal structures presented in this study.