Ipomoelin, a Jacalin-Related Lectin with a Compact Tetrameric Association and Versatile Carbohydrate Binding Properties Regulated by Its N Terminus

Many proteins are induced in the plant defense response to biotic stress or mechanical wounding. One group is lectins. Ipomoelin (IPO) is one of the wound-inducible proteins of sweet potato (Ipomoea batatas cv. Tainung 57) and is a Jacalin-related lectin (JRL). In this study, we resolved the crystal structures of IPO in its apo form and in complex with carbohydrates such as methyl α-D-mannopyranoside (Me-Man), methyl α-D-glucopyranoside (Me-Glc), and methyl α-D-galactopyranoside (Me-Gal) in different space groups. The packing diagrams indicated that IPO might represent a compact tetrameric association in the JRL family. The protomer of IPO showed a canonical β-prism fold with 12 strands of β-sheets but with 2 additional short β-strands at the N terminus. A truncated IPO (ΔN10IPO) by removing the 2 short β-strands of the N terminus was used to reveal its role in a tetrameric association. Gel filtration chromatography confirmed IPO as a tetrameric form in solution. Isothermal titration calorimetry determined the binding constants (KA) of IPO and ΔN10IPO against various carbohydrates. IPO could bind to Me-Man, Me-Glc, and Me-Gal with similar binding constants. In contrast, ΔN10IPO showed high binding ability to Me-Man and Me-Glc but could not bind to Me-Gal. Our structural and functional analysis of IPO revealed that its compact tetrameric association and carbohydrate binding polyspecificity could be regulated by the 2 additional N-terminal β-strands. The versatile carbohydrate binding properties of IPO might play a role in plant defense.


Introduction
Plant defense is a complicated mechanism in response to mechanical wounding, herbivore and microorganism attack. Many proteins, namely wound-inducible proteins, are expressed to prevent pathogen infection, inhibit digestion by insects, and repair injured tissues [1,2]. One group of wound-inducible proteins is lectin, the carbohydrate binding protein [3,4]. Plant lectins are involved in the plant defense mechanism because of carbohydrate binding properties [5][6][7][8][9]. The toxicity of lectins was also confirmed in animal experiments [10,11]. Plant lectins show resistance to digestive enzymes and can bind selectively to the carbohydrate moieties of gut epithelial cells to interfere in nutrient digestion and absorption [12], so they could be a natural insecticide. In addition, plant lectins have been used for blood typing and immunological assay. The lectin concanavalin A is commercially used in affinity chromatography for purifying glycoproteins. Plant lectins have long been reported as potential inhibitors of viruses [13][14][15][16][17].
Most plant lectins were originally isolated from seeds and vegetative storage tissues. Accumulating data have revealed that plants ubiquitously synthesize lectins in response to abiotic and biotic stresses. These inducible lectins are synthesized and then exported to vacuoles by signal peptides or reside in the cytoplasm [18,19]. The physiological function of plant lectins for subcellular localization remains obscure. However, the major assumption is that lectins are involved in defense and may also have a role in signal transduction for response to stress [20]. Structure analysis of plant lectins demonstrated a diverse group of proteins that can be classified into 6 different groups (http:// www.cermav.cnrs.fr/lectines/): monocot lectin, hevein domain lectins, b-prism lectins, b-trefoil lectins, cyanovirin-N homologs, and legume lectin.
Jacalin-related lectins (JRL) have a b-prism fold. In 1996, the structure of Jacalin from seed of jackfruit (Artocarpus integrifolia) was first reported to have a tetrameric association for binding to galactose [21]. Later, Maclura pomifera seed agglutinin was reported to have the same tetrameric structure as Jacalin [22]. The other lectin, Artocarpin, from seed of jackfruit (Artocarpus heterophyllus) shares the same tetrameric association for binding to mannose [23]. Moringa M from black mulberry (Morus nigra) forms a tetrameric association like that of Jacalin [24]. JRLs were once thought to be confined to the Moraceae. However, increasing structural evidence reveals that the lectins with a bprism fold exist universally in plants and animals [25] but with different quaternary association. Heltuba is a plant tuber lectin from Helianthus tuberosus (Jerusalem artichoke) that has a donut shape with an octahedral assembly by the b-prism building block [26]. Caselpa is a rhizome lectin from Calystegia sepium (Hedge bindweed) that has a dimeric form [27]. PPL is a plant seed lectin from Parkia platycephala that contains 3 repetitive bprism domains and forms a dimeric form with hexahedral assembly [28].
Ipomoelin (IPO), expressed in the leaves of sweet potato (Ipomoea batatas cv. Tainung 57), was found easily inducible by wounding and methyl jasmonate [29,30]. Previous study showed that IPO can agglutinate human blood and bind to different carbohydrates, such as methyl a-D-mannopyranoside (Me-Man), methyl a-Dglucopyranoside (Me-Glc), mannose, glucose and galactose [10]. In this study, we resolved the crystal structures of IPO in the apo form and in complex with Me-Man, Me-Glc and methyl a-Dgalactopyranoside (Me-Gal) to reveal the different quaternary associations of IPO and its binding pocket for carbohydrates. A The packing diagram of apo IPO with 5 molecules in green. Four of 5 molecules form a tetramer and the 5th molecule can form another tetramer by the red one, the yellow one and the blue one in the center. The molecules in red, yellow and blue are generated by symmetric operations (-X, Y, -Z), (X, -Y, -Z), and (-X-1, -Y, Z). (B) The resolved IPO-Me-Man complex in green is 2 molecules in an asymmetric unit. However, the other 2 molecules in red are generated by the symmetric operation (X, -Y, -Z) to form a tetramer in the center. (C) The resolved IPO-Me-Glc complex is a tetrameric form, and (D) the IPO-Me-Gal complex is also a tetramer. All packing diagrams reveal its tetrameric nature. doi:10.1371/journal.pone.0040618.g001 The carbohydrate binding pockets are indicated by green mesh. Tetrameric IPO is represented by monomer A in blue, monomer B in purple, monomer C in light blue, and monomer D in pink. The symmetric axis is represented by a black ellipse in the center of tetramer. (C) Structure-based multiple sequence alignment of Jacalin family. Five homologs were selected for sequence comparison from resolved protein structures: Ipomoelin-tetramer from Ipomoea batatas (PDB: 3R52); Calsepa-dimer from Calystegia sepium (PDB: 1OUW); Banlec-dimer from Musa acuminate (PDB: 2BMZ); Jacalin-tetramer from Artocarpus hirsutus (PDB: 1TOQ); Parkia-hexamer from Parkia platycephala (PDB: 1ZGS); and Heltuba-octomer from Helianthus tuberosus (PDB: 1C3K). Positions of identical conserved residues are shown in white on dark grey background, and regions of similarly conserved residues in light grey are boxed. Representation of secondary structure elements and numbering above the alignment is based on the IPO structure. The secondary structure elements below the alignment are based on the Heltuba truncated IPO (DN10IPO) was prepared to reveal its role in tetrameric association in solution by gel filtration chromatography. In addition, the carbohydrate binding constants of IPO and DN10IPO were determined by isothermal titration calorimetry (ITC). DN10IPO showed a recovered mannose/glucose-specific lectin. Structural and functional analysis identified IPO as a member of the JRL family but with a different tetrameric association. The N-terminus of IPO plays a critical role in regulating broad carbohydrate binding.

Crystal Packings of Apo IPO and IPO-carbohydrate Complexes Show Tetrameric Association
The apo IPO showed an orthorhombic space group of I222. A reasonable volume of the unit cell (Vm) for the Matthew coefficient was estimated at 2.19 Å 3 /Da and 44% solvent content by 8 IPO molecules. However, only 5 IPO molecules in an asymmetric unit could be built after molecular replacement. We structure. The carbohydrates Me-Man, Me-Glc, and Me-Gal share 9 hydrogen-bonding interactions with Gly21, Tyr97, Gly141, Trp142, Tyr143 and Asp145 of IPO (blue triangle). The residues of IPO located at the interface are boxed in red. The two short b strands at the N terminus are also involved in the interface. The underlined Jacalin-tetramer representing the sequence is extracted from the C terminus of Jacalin (chain B). doi:10.1371/journal.pone.0040618.g002 obtained a higher Matthew coefficient with 3.51 Å 3 /Da and 65% solvent content. In the packing diagram for apo IPO, we observed a tetrameric association with an additional monomer in an asymmetric unit ( Figure 1A). The additional monomer could form a tetrameric association with the other 3 neighboring molecules, which were generated by symmetric operations (-X, Y, -Z), (X, -Y, -Z), and (-X-1, -Y, Z). So 4 IPO molecules could form a tetramer.
To determine the carbohydrate binding pocket of IPO, carbohydrates such as Me-Man, Me-Glc and Me-Gal were used to co-crystallize with the IPO protein. The crystals of IPOcarbohydrate complexes were determined in different space groups. IPO-Me-Man belongs to an orthorhombic space group C222 1 . The Matthew coefficient and solvent content for IPO-Me-Man had a reasonable value of 2.21 Å 3 /Da and 44.4% for 2 molecules in an asymmetric unit. Although only 2 IPO molecules were built in the IPO-Me-Man complex, the other 2 IPO molecules could be generated by symmetric operation (X, -Y, -Z) and resulted in a tetrameric association ( Figure 1B). The crystal of IPO-Me-Glc was determined to be a monoclinic space group P2 1 . The Matthews coefficient and solvent content was 2.26 Å 3 /Da and 45.5% for 4 molecules. The packing results for IPO-Me-Man and IPO-Me-Glc indicated that the carbohydrates binding to IPO might result in a compact packing as compared with that of apo IPO. In addition, the resolved structure of IPO-Me-Glc formed a tetrameric association ( Figure 1C).
IPO-Me-Gal belongs to an orthorhombic space group P2 1 2 1 2 1 . The Matthews coefficient and solvent content were 2.25 Å 3 /Da and 45.2%, respectively, for 4 molecules in an asymmetric unit. The 4 IPO-Me-Gal molecules shown in Figure 1D form the same tetrameric association as that of IPO-Me-Glc. On the basis of crystal packings of apo IPO and IPO-carbohydrate complexes, IPO would form a tetrameric association.

Overall Structure of Monomeric IPO and its Tetrameric Association
The monomeric IPO from residues 1 to 154 shows a typical bprism fold found in the JRL family, with 12 b-sheets (b3-b14) and 2 additional short, extended, N-terminal b-strands (b1-b2) (Figure 2A and 2C). Each b-prism fold comprises 3 Greek-key motifs forming 3 planes by 3 four-stranded b-sheets: plane 1 by b3 to b4 and b13 to b14; plane 2 by b5 to b8; plane 3 by b9 to b12.
Four IPO protomers form a compact tetrameric association by swapping their extended N termini from residues 1 to 10. We analyzed the tetrameric association of IPO-Me-Glc. As shown in Figure 2B, the 2 extended N termini from monomer A in blue and monomer B in purple swap with each other. The interacting interface between the four IPO protomers is formed by the extended N termini. Consequently, a larger buried interface between monomers A and B is 1,522 Å 2 . The residues located at the interface are 2-10, 12, 15-30, 59-67, 91-92, 98, 121, 134, 137, 139-140, 146, 150, and 152 in monomer A (as shown in red box in Figure 2C). In total, 13 hydrogen bonds are formed by the residues Leu5, His8, Asn19, Gln22, Ser25, Arg27, Asp60, Ile61, Thr63, Thr121, Asn139 and Tyr150 in the interface between monomers A and B. The buried interface between monomer C and monomer D is 1,554 Å 2 . Furthermore, the buried interface between monomers A and C is 755 Å 2 , which is mainly contributed by the interacting residues of N-terminal residues 4 to 17 and C-terminal residues 91, 121-126, 128, and 151. In addition, the interface between monomers D and B is 731 Å 2 .

The Carbohydrate Binding Pocket of IPO
The carbohydrate binding pocket of IPO was confirmed at loops b13 and b14 by the structures of IPO-Me-Glc, IPO-Me-Man and IPO-Me-Gal (as shown in Figure 2B with green mesh). In the chain A of IPO-Me-Glc, 9 hydrogen bonds are formed by the residues Gly21, Tyr97, Gly141, Trp142, Tyr143 and Asp145 of IPO and the atoms O1, O3, O4, O5, and O6 of Me-Glc ( Figure 3A and Table 1). The atom C7 of Me-Glc is involved in the methyl carbon (Me)…p interaction with Trp142 of IPO. The hydrogen bonds are slightly different between chain A and chains B to D. The hydrogen bonds of chains B to D are formed between the same residues of chain A and Me-Man, except for   Table 1). The differences might result from the binding of cadmium ion (Cd 2+ ). In chains B to D, the Cd 2+ atom forms 5 coordinates by the O atom of the carbonyl group of Asn19, OG atom of Ser18, and 3 water molecules. One of the 3 water molecules forms a hydrogen bond with Asp145 ( Figure 3B).
In the structure of IPO-Me-Man, two IPO protomers were built, and only one Me-Man molecule could be observed in chain A. The temperature factor of Me-Man in the structure of IPO-Me-Man is 66.5 Å 2 , which is higher than that of Me-Glc, with 34.5 Å 2 (Table 1). This phenomenon might indicate that only a few Me-Man molecules bound to IPO proteins in IPO-Me-Man, which resulted in a higher temperature factor. Nine hydrogen bonds are formed by the residues Gly21, Tyr97, Gly141, Trp142, Tyr143, and Asp145 of IPO and the atoms O1, O3, O4, O5 and O6 of Me-Man ( Figure 3C and Table 1). The atom C7 of Me-Man is also involved in the Me…p interaction with Trp142 of IPO. The binding orientation of Me-Man is similar to that of Me-Glc. In the structure IPO-Me-Gal, 10 hydrogen bonds are formed by the same residues Gly21, Tyr97, Gly141, Trp142, Tyr143, and Asp145 of IPO ( Figure 3D and Table 1). The atom C7 of Me-Gal is shown in the Me…p interaction with Trp142 of IPO. This revealed the importance of the methyl group of carbohydrates for binding to IPO.

The Tetrameric Form of IPO Identified by Gel Filtration Chromatography
To validate that the quaternary association of IPO is also a tetrameric form in solution, purified IPO was used in gel filtration experiments. The molecular mass of IPO could be calculated according to the linear regression equation of the standard protein markers purchased from BioRad ( Figure 4C). In the preliminary study, IPO protein was dissolved in running buffer (27 mM Tris-HCl pH 7.0, 2 M NaCl) without additional carbohydrates. We obtained a retarded result, with corresponding molecular mass 4.0 kDa (Peak 3 in Figure 4A). Thus, IPO has the binding ability of dextran in the matrix of the Superdex 200 column. To eliminate the binding effect of IPO to dextran, running buffer was prepared with an additional 0.2 M Me-Glc, and a shift of the IPO peak could be observed, with corresponding molecular mass of 53.3 kDa (Peak 2 in Figure 4A). Consequently, running buffer with an additional 1 M glucose was prepared to totally eliminate the binding effect of IPO. The corresponding molecular mass of IPO in solution was 64.7 kDa (Peak 1 in Figure 4A). The molecular mass of recombinant IPO with a His tag was 17.3 kDa for a monomer and 69.2 kDa for a tetramer. The results from gel filtration experiments demonstrated that IPO shows a tetrameric association in solution.
To further identify the role of the N terminus in the tetramerization of IPO, we prepared a truncated IPO (DN10IPO) by removing residues 1 to 10 to monitor the change in quaternary association. The native IPO protein or the truncated IPO protein was dissolved in the running buffer with 1 M glucose. Peak 1 in Figure 4B represents the native IPO, with molecular mass 63.2 kDa, which is a tetrameric size. Peak 2 in Figure 4B represents the DN10IPO, with molecular mass 21.9 kDa, which is near the truncated monomer size (16.3 kDa). The results further confirmed that IPO has a tetrameric association and its N terminus plays an important role in forming a tetramer.

Binding Constants of IPO and Truncated IPO to Various Carbohydrates Detected by ITC
To determine the binding constants of IPO to Me-Man, Me-Glc and Me-Gal, 1 mM IPO solution was titrated with 25 mM carbohydrate solution. The interaction of IPO and carbohydrate was an exothermal reaction. The optimal curves and thermodynamics parameters could be fitting and calculated by Microcal Origin 7.0. The K A of IPO to Me-Man was the highest, 7.04610 3 M 21 . The K A values for Me-Gal and Me-Glc were 4.09610 3 M 21 and 2.01610 3 M 21 , respectively (Table 2 and Figure 5).
Subsequently, carbohydrates without the methyl group were used to determine the binding affinity of IPO. From preliminary study, 1 mM IPO titrated with 25 mM Man, Glc, and Gal revealed no obvious exothermal reaction. After increasing the concentration with 3 mM IPO titrated with 75 mM Man, Glc, and Gal, the exothermal curves could be observed and calculated. The K A values for IPO binding to Man, Gal and Glc were 1.05610 2 M 21 , 0.57610 2 M 21 , and 0.32610 2 M 21 (Table 2 and Figure 5). Thus, the interactions between IPO and carbohydrates were stronger with than without the methyl group.  Figure 6). Thus, the N-terminus of IPO is involved in tetramerization in regulating the binding affinity to carbohydrates.

Various Quaternary Structures in the JRL Family
We submitted the coordinates of a monomer of apo IPO (e.g., chain A; Figure S1C) to the web service Matras for 3-D protein structure comparison [31]. We found the highest Z-score, 124.5, for the template structure, a dimeric form of Calsepa from Calydyrgia sepium (PDB: 1OUW; Figure S1D) [27], in our molecular replacement procedure. The following structures were PPL from Parkia platycephala with a hexahedral ring (PDB: 1ZGR; Figure  S1E) [28], Heltuba from Helianthus tuberosus with an octahedral ring  Figure S1F) [26], Banlec from banana with an another kind of dimeric form (PDB: 2BMZ; Figure S1B) [32], and Jacalin from jackfruit seeds with a tetrameric form (PDB:1UGW; Figure S1A) [33]. These data indicate the various quaternary structures in the JRL family, despite the same b-prism fold of protomer.
The various quaternary associations in the JRL family exhibited different contacts between protomers. A previous report indicated that the buried interface of the Calsepa dimer is 1,327 Å 2 by a probe with 1.6 Å radius [27]. Here, we analyzed the buried interface of the selected structures from the above comparison by using the PDBe PISA service with 1.4 Å radius [34]. The buried interface area from tetrameric IPO encompasses 1,539 Å 2 , which is larger than that of Calsepa (1,202 Å 2 ), PPL (1,294 Å 2 ), Banlec (750 Å 2 ), Heltuba (736 Å 2 ), and Jacalin (1023 Å 2 ). The N terminus of the protomer in the JRL family has an important role in the quaternary association by swapping in the interface and then forming a dimer, tetramer, hexamer, and octomer. To compare the difference between the tetrameric Jacalin ( Figure  S1A) and the tetrameric IPO ( Figure S1C), the tetramer of Jacalin showed a looser interface than that of IPO. Therefore, IPO formed a different compact tetramer.

The Carbohydrate Binding Pocket of IPO Reveals its Versatile Binding Properties
In this study, we resolved the crystal structures of IPO-Me-Man, IPO-Me-Glc and IPO-Me-Gal complexes. These monosaccharides showed similar orientation to bind to IPO. The binding pocket of IPO contains 6 residues such as Gly21, Tyr97, Gly141, Trp142, Tyr143 and Asp145, to form hydrogen bonds with different monosaccharides (Figure 3 Table 2). The carbohydrate binding manner of IPO is not confined as is the mannose-glucose-specific binding lectin.
In addition to determining monosaccharides with the methyl group, we used monosaccharides without a methyl group, such as mannose (Man), glucose (Glc), and galactose (Gal), to determine their binding constant to IPO. Since the lower binding affinity of IPO titrated with Man, Glc or Gal couldn't get the best fitting for the titration curves, the n value was consequently fixed at 1.0 for fitting the curves (  [36]. The results show no differences with or without the methyl group of monosaccharides for binding properties in Artocarpin and Banlec possibly because of no aromatic side chain of residues in Artocarpin and Banlec like the residue Trp142 in IPO ( Figure 7A and 7B). Interestingly, IPO shared similar binding properties to Jacalin for its Tyr122, which  [37]. To examine the binding mode of Me-Man for Artocarpin, Banlec, Jacalin, and IPO, the binding position of Me-Man with IPO showed a distant binding site as compared with that for Artocarpin, Banlec, and Jacalin ( Figure 7D). DN10IPO could be recovered as the mannose/glucose specific lectin if DN10IPO represented the monomeric IPO and wild-type IPO represented the tetrameric IPO. The monomeric IPO showed 5 times and 6 times binding affinity to Me-Man and Me-Glc, respectively, as compared with those of tetrameric IPO. Therefore, the N terminus of IPO is involved in the carbohydrate recognition, which results in the carbohydrate binding polyspecificity of tetrameric IPO. From the tetrameric IPO structure, the residue Leu5 and His8 in the N terminus of monomer B (chain B) forms 3 hydrogen bonds with the residue Asn19 in the loop between b3 and b4 of monomer A (chain A) (Figure 8). The hydrogen bonds might pull out the loop of b3-b4 and form a larger binding cavity for different carbohydrates in monomer A. However, in DN10IPO, the hydrogen bonds would disappear and might relocate the b3-b4 loop to cause a smaller binding cavity. The axial O4 of Me-Gal would not easily enter into the smaller binding cavity. The results might be confirmed by the crystal structure of DN10IPO-Me-Man in further study.

The N-terminus of IPO is Involved in Tetramerization and Regulates the Carbohydrate Binding Specificity
In conclusion, we resolved the structures of apo IPO and IPO in complex with Me-Man, Me-Glc and Me-Gal. IPO is proposed to have a tetrameric association by 4 protomers of the b-prism with an additional N terminus, which shows a compact tetrameric association in the JRL family. From gel filtration experiments, we confirmed the tetrameric association of IPO in solution. The N terminus of IPO plays an important role in forming a tetramer. In addition, the binding pocket of IPO was identified and found to bind to Me-Glc, Me-Man, and Me-Gal with similar hydrogen bond networks. Furthermore, the binding constants of IPO were determined by ITC. The IPO structures further extend the diverse quaternary structures of the JRL family of plants and show versatile carbohydrate binding properties regulated by the N terminus. Thus, the wound-inducible protein IPO from sweet potato has versatile carbohydrate binding properties and might play a role in plant defense.

Protein Expression and Purification
The pTZ18UH-IPO and pTZ18UH-DN10IPO vectors were transformed into Escherichia coli BL21 (DE3) cells (Novagen). A single colony was cultured in 5 ml LB medium containing 100 mg/ ml ampicillin (LB/Amp) at 37uC overnight. The medium was further transferred into 600 ml LB/Amp to an A 600 of about 0.5 to 0.7 and then induced with 0.1 mM isopropyl-b-D-thiogalactopyranoside (IPTG) at 25uC for 6 hr. Cells harvested by centrifugation were resuspended in a loading buffer (20 mM sodium phosphate, pH 7.4, 0.5 M sodium chloride, 20 mM imidazole). After breaking cells by use of an ultrasonicator (Sonicator 3000, Misonix), the supernatant of the crude cell lysate was loaded onto a Histrap FF column (GE Healthcare) with use of an Ä kta Prime fast protein liquid chromatography (FPLC) system (GE Healthcare). After washing the Histrap FF column with 3x column volume of loading buffer (1x phosphate buffered saline, 5 mM adenosine triphosphate, 10 mM MgSO 4 ), the IPO protein was eluted by use of elution buffer (50 mM sodium phosphate, pH 7.4,  Table 3. Crystallography statistics for apo ipomoelin (IPO) and IPO in complex with carbohydrates methyl a-D-mannopyranoside (Me-Man), methyl a-D-glucopyranoside (Me-Glc) and methyl a-D-galactopyranoside (Me-Gal). sodium formate, 40% w/v polyethylene glycol 3,350. The crystals of IPO-Me-Gal appeared in 7 days. A mixture of the reservoir solution with 100% glycerol in a 4:1 volume ratio was used as cryo-protectant for data collection. The diffraction data were collected at 100K and detected by a Quantum 315 or Quantum 210 CCD detector at the BL13B1 or BL13C1 beamlines of NSRRC (Hsinchu, Taiwan). All diffraction data were processed and scaled with use of the HKL2000 program [38]. The diffraction statistics are in Table 3.

Structure Determination and Refinement
We used a blastp search for the amino acid sequence of IPO [GenBank: BAA14024.1] against the algorithm of the National Center for Biotechnology Information (NCBI) protein databank database for searching structural templates. The amino acid sequence of Calystegia sepium agglutinin (Calsepa), a JRL (PDB: 1OUW), showed 53% sequence identity to that of IPO. The monomeric structure of Calsepa was further used in a search to determine the structure of apo IPO by molecular replacement with use of the program CNS [39]. After cross-rotation and translation of molecular replacement, 4 values were obtained. Initial rigid body refinement for the 4 monomeric structures gave a 48.8% R-factor. Clear continuous electron density could be observed after calculation of Fourier maps, and the 5 th molecule of apo IPO was further built accordingly. Because of different space groups for the structures of the IPO-Me-Man, IPO-Me-Glc and IPO-Me-Gal complexes, the resolved monomeric apo IPO was used as a search template in the following molecular replacement method. The solutions of cross-rotation and translation could be obtained with 2 molecules for the IPO-Me-Man complex, 4 molecules for IPO-Me-Glc and 4 molecules for IPO-Me-Gal. Those solutions were further applied to initial rigid body refinement, and reasonable values were obtained (e.g., 36.7% Rfactor for IPO-Me-Man, 35.4% for IPO-Me-Glc, and 32.9% for IPO-Me-Gal).
Manual model rebuilding involved use of Coot [40], alternating refinement by the CNS program, with 5% or 10% of the observed reflections randomly selected and set aside for calculation of the R free value. The final refined statistics are in Table 3. For the protein interface of the tetrameric form, IPO-Me-Glc was used as a representative for analysis by the web service PDBe PISA [34]. All molecular representations were prepared with use of Deep-View [41] and PyMOL [42]. The coordinates of monomers of apo IPO (e.g., chain A) were subjected to the web service Matras for structure comparison [31].

Quantification of Protein and Carbohydrate Solution for Binding Assay
The quantification of protein solution for binding assay was determined by the UV absorption method. The purified IPO protein was dialyzed against 20 mM Tris-HCl, 150 mM NaCl (pH 7.0) at 4uC overnight. The concentration of IPO and DN10IPO was determined by UV absorption spectroscopy at 280 nm with the specific extinction coefficient e of 22,920 M 21 cm 21 , which was determined from the prediction of IPO primary sequence. From Beer-Lambert law, A = e6b6C where A is the absorbance of the sample at 280 nm, b is the pathlength in 1 cm, and C is the protein concentration (M).
The protein concentration C could be calculated from the equation. Carbohydrates were prepared by weighting the amount on a microbalance before dissolving in dialysis buffer (20 mM Tris-HCl, 150 mM NaCl pH 7.0).

Binding Affinity by Isothermal Titration Calorimetry (ITC)
ITC measurements involved use of a MicroCal iTC200 microcalorimeter (GE Healthcare) at 25uC. In individual titration, 1-2 ml carbohydrate solution was added at 180-s intervals by use of a computer-controlled 40 ml syringe to a cell containing 280 ml IPO protein solution under constant stirring at 1,000 rpm. The concentration of IPO protein was 1-3 mM and that of Me-Man, Me-Glc, Me-Gal, Man, Glc and Gal 25-75 mM. The titration of carbohydrate solution in this range of concentration to the dialysis buffer was used as a control. Measurements of the heat change determined from the binding constant (K A ), reaction stoichiometry (n), and enthalpy (DH). The 18 experimental data were fitted for a 1:1 binding model (one-site of fitting) with Microcal Origin 7.0 software. Free energy (DG) and binding entropy (DS) were calculated by the equations DG = -RTlnK A and DG = DH -TDS. R is the gas constant and T the absolute temperature. The optimal c-value in ITC calculation varied between 1 and 10. However, for titrations with Man, Glc and Gal, the c-values were ,1.

Protein Data Bank Accession Codes
The atomic coordinates and structure factors of apo IPO and IPO-carbohydrate structures have been deposited in the RCSB Protein Data bank, with 3R50 for apo IPO, 3R51 for IPO-Me-Man complex, 3R52 for IPO-Me-Glc complex and 4DDN for IPO-Me-Gal complex.