2.1. Cloning, expression and purification

cDNA coding for the endo-1,3--glucanase from S. tuberosum was amplified by PCR using the gluB20-2 ORF and cloned into pTOPO vector as described previously (Wojtkowiak et al., 2012). The E259A replacement was introduced into the coding sequence of gluB20-2 by site-directed mutagenesis following protocols described by Green & Sambrook (2012), with the use of forward (GTATCGGCAAGTGGTTGGCCTTCTGAGGGACA) and reverse (GGTGTCCCTCAGAAGGCCAACCACTTGCCGAT) primers coding for the mutation (bold). The mutant cDNA was cloned into pET30a vector and sequenced. The recombinant gene was expressed in Escherichia coli BL21 strain (Studier & Moffatt, 1986) using TB medium (Green & Sambrook, 2012). The bacterial culture was incubated at 310 K until its optical density reached A₆₀₀ = 1. Protein synthesis was induced by the addition of isopropyl -D-1-thiogalactopyranoside (IPTG) to a final concentration of 0.25 mM and incubation was continued at 291 K for 24 h. Finally, the culture was centrifuged at 8000g at room temperature for 10 min. The bacterial cells were collected and kept overnight at 253 K. Protein purification was performed as described by Wojtkowiak et al. (2012). The homogenous protein fractions were pooled together, dialyzed against 25 mM sodium phosphate buffer pH 7.8 with the addition of 10%(v/v) glycerol and kept as 1 mg aliquots at 253 K. For crystallization experiments, the protein was concentrated to 7 mg ml^-1 and the buffer was exchanged to 20 mM Tris-HCl pH 8.0 using Millipore Centricon 10 filters.

2.2. Crystallization

Prior to crystallization experiments, the E259A mutant of recombinant potato endo-1,3--glucanase was incubated for 3 h at 277 K with either a hexameric or a pentameric linear (13)--glucan oligosaccharide. The hexasaccharides and pentasaccharides, which were derived from longer (13)--glucan polymers and isolated as high-purity samples, were a kind gift from Professor Tadeusz Antczak (Technical University of Lodz). Sparse-matrix screening (Structure Screen 1 and 2; Molecular Dimensions Ltd) for crystallization conditions of a pentasaccharide complex was not successful, whereas the incubation mixture with the hexasaccharide yielded crystals of two different complexes with the protein. In one case (3G) crystals were obtained with a threefold molar excess of the oligosaccharide and in the other case (7G) with a sixfold molar excess of the ligand. Crystals of both complexes were obtained by vapour diffusion at 292 K in hanging drops made up of 1.5 µl protein/ligand solution and 1.5 µl reservoir solution consisting of 0.1 M sodium citrate pH 5.6, 0.2 M ammonium acetate, 24% PEG 8000. Crystals grew over a period of 2-7 d in the form of bunched needles. Single and larger crystals were obtained by the use of streak-seeding and were left in the crystallization drops until data collection (two months).

2.3. Data collection and processing

X-ray diffraction data for both complexes were collected on beamline X11 at the EMBL Hamburg with 0.5° oscillation and 1.68 Å resolution for the 3G complex and with 0.75° oscillation and 1.55 Å resolution for the 7G complex. The data collection was performed at 100 K with the use of 40%(v/v) PEG 400 mixed in a 1:1(v:v) ratio with the reservoir solution as a cryoprotectant. The data were indexed, integrated and scaled with HKL-2000 (Otwinowski & Minor, 1997). In both cases the crystals were monoclinic, belonging to space group P2₁, and the asymmetric unit contained one protein molecule. A summary of the data-collection and processing statistics is given in Table 1.

Table 1 X-ray data-collection and refinement statistics Values in parentheses are for the highest resolution shell.   3G 7G Data collection   Radiation source X11, EMBL Hamburg X11, EMBL Hamburg   Wavelength (Å) 0.81620 0.81480   Temperature of measurements (K) 100 100   Space group P21 P21   Unit-cell parameters (Å, °) a = 54.9, b = 49.3, c = 57.1, = 98.4 a = 55.3, b = 49.2, c = 57.4, = 98.8   Mosaicity (°) 0.781 0.488   Protein molecules in asymmetric unit 1 1   Solvent content (%) 41 42   Resolution range (Å) 30.00-1.68 (1.74-1.68) 40.00-1.55 (1.61-1.55)   No. of unique reflections 34231 44391   Rmerge# 0.069 (0.264) 0.091 (0.680)   Completeness (%) 99.0 (92.7) 100.0 (100.0)   <I/(I)> 23.2 (5.1) 15.3 (2.1)   Multiplicity 6.2 (4.8) 3.8 (3.8) Refinement statistics   Resolution (Å) 20.0-1.68 20.0-1.55   No. of reflections 34202 44352   No. of reflections in test set 1076 1139   Rwork+ 0.136 0.151   Rfree 0.171 0.181   No. of residues 315 315   No. of non-H atoms     Protein 2524 2528     Ligand 34 57     Water molecules 338 356   R.m.s.d. from ideal bond lengths (Å) 0.017 0.017   R.m.s.d. from ideal bond angles (°) 1.57 1.64   Average B factors (Å2) 17.5 15.3   Ramachandran statistics (%)     Most favoured regions 91.1 90.4     Additionally allowed regions 8.9 9.6   PDB code 4gzi 4gzj #Rmerge = , where Ii(hkl) is the intensity of observation i of reflection hkl. +Rwork = for reflections in the working set, where Fobs and Fcalc are the observed and calculated structure factors, respectively. Rfree is calculated analogously for test reflections randomly selected and excluded from the refinement.

2.4. Structure determination and refinement

The structure of the 3G complex was determined by molecular replacement using the coordinates of native potato endo-1,3--glucanase (PDB entry 3ur7 , molecule A; Wojtkowiak et al., 2012) as a search model. The rotational and translational searches were performed with the MOLREP program (Vagin & Teplyakov, 2010). Structural refinement was carried out in REFMAC5 (Murshudov et al., 2011) using maximum-likelihood targets. Manual adjustments of the model in electron-density maps were performed in the XtalView program (McRee, 1999). For TLS analysis of the protein model, the TLSMD server was used (Painter & Merritt, 2006) and three TLS segments were defined per polypeptide chain. When the refinement reached R and R_free values of 0.243 and 0.276, respectively, a difference map was calculated, which unambiguously showed a trisaccharide (laminaratriose) molecule docked in one-half of the binding cleft. The ligand molecule was built manually according to the electron density with the aid of the PRODRG server (Schüttelkopf & van Aalten, 2004). In the final stages of modelling water molecules were added to the coordinate set and the refinement converged at R and R_free values of 0.136 and 0.171, respectively, with the use of all reflections to the full resolution of 1.68 Å. The 3G structure was subsequently used for the determination of the structure of the 7G complex. All structure-determination and refinement steps for the 7G complex were as for the 3G complex, except that the difference electron-density maps showed not only the same trisaccharide as before but also another oligosaccharide bound in the second half of the catalytic cleft. From the space spanned by this ligand it is obvious that it consists of four glucose units (laminaratetrose), but only two of them (at subsites +3 and +4) could be built in electron-density maps without ambiguity. The refinement of the 7G structure converged at R and R_free values of 0.151 and 0.181, respectively, for all reflections to the full resolution of 1.55 Å. A summary of the refinement statistics is given in Table 1. Structural superpositions were calculated using ALIGN (Cohen, 1997). Molecular and electron-density illustrations were prepared in PyMOL v.1.3 (Schrödinger LLC).

2.5. Mass-spectrometry experiments

The molecular masses of the oligosaccharides were determinated using a Kratos Kompact-SEQ MALDI-TOF mass spectrometer. Prior to analysis, the samples were purified by the use of Dowex 50WX8 ion-exchange resin. The molecular masses were estimated from the m/z ratio of the ionic adducts (M + Na⁺). 2,5-Dihydroxybenzoic acid and 2',4',6'-trihydroxyacetophenone were used as a matrix.

3.1. Overall structures

Crystalline complexes of E259A GLUB20-2 with oligosaccharides were obtained by cocrystallization. For the cocrystallization experiments, a pentasaccharide or a hexasaccharide derived from (13)--glucan (laminaran) were used as potential ligands. In the case of the pentasaccharide, all crystallization trials failed. In contrast, in the case of the hexasaccharide ligand two different complexes were crystallized. In one case, which was obtained with a lower protein:hexasaccharide molar ratio (1:3), the protein subsites -1 to -3 were occupied by a trisaccharide (laminaratriose). In the other crystal, which was grown  at a higher molar excess of the ligand (1:6), in addition to the trisaccharide found in exactly the same place an additional tetrasaccharide (laminaratetrose) was bound at subsites +1 to +4. The tetrasaccharide was less perfectly bound in its binding site and only the glucose units at subsites +3 and +4 were included in the model (at full occupancy). However, the distinct electron density that extends from subsite +3 towards the catalytic centre (especially at +1) was interpreted as indicating that a tetrasaccharide was present at this site. Indeed, the available space is perfectly compatible with four glucose units and a provisionally docked tetrasaccharide at this site would have its nonreducing end within 2.9 Å of the reducing end of the fully modelled trisaccharide. Moreover, the presence of a tetrasaccharide was confirmed experimentally by mass-spectrometric analysis of an incubation buffer that simulated the crystallization conditions (see below). However, because of the evident disorder at subsites +1 and +2, the tetrasaccharide model included in the coordinate set is limited to residues +3 and +4 which could be modelled with confidence. According to the number of occupied subsites, the complex structures are denoted as 3G and 7G in the following.

Both complexes crystallized in space group P2₁ with one protein molecule in the asymmetric unit and with very similar unit-cell parameters (Table 1). The 3G and 7G crystal structures were determined by molecular replacement and refined to 1.68 and 1.55 Å resolution, respectively, with good refinement statistics (Table 1). The models of both structures contained all 315 amino-acid residues of the mature protein (residues 24-338). The C-terminal histidine tag (residues 339-346) could not be modelled owing to disorder. It is interesting to note that in the crystal structures of the ligand-free native protein reported previously the His tag was docked into the active site of an adjacent protein molecule, linking them into infinite linear chains (Wojtkowiak et al., 2012).

The trisaccharide molecule is docked at the -1, -2 and -3 subsites, with its nonreducing terminus at the far end of the binding cleft (-3) marked by the 2-2 loop of the protein. The tetrasaccharide is bound at the plus-numbered subsites with its reducing terminus at the other end of the binding cleft (+4) near helix 6. In this binding mode, the reducing end at subsite -1 and the nonreducing end at subsite +1 meet at the catalytic centre, showing the pattern of products of the hydrolytic reaction catalyzed by the enzyme. At the same time, the scheme of interactions of the polar groups of the saccharides with the protein residues defines the precise enzyme-substrate recognition pattern in atomic detail.

3.2. Structural comparisons

The protein molecules in both structures are essentially identical, as illustrated by the small r.m.s.d. value (0.12 Å) calculated for all their C atoms. The trisaccharide molecules bound in the catalytic clefts of both structures occupy the same position and have an identical conformation, as illustrated by the r.m.s.d. value of 0.09 Å calculated for all of their non-H atoms in a superposition based on the protein C atoms only. Superpositions of the present mutant structures with the native protein molecule (molecule A from PDB entry 3ur7 ; Wojtkowiak et al., 2012) also yielded very small r.m.s.d. values for all C atoms (0.22 and 0.25 Å for 3G and 7G, respectively), demonstrating an identical overall conformation. The E259A GLUB20-2 molecule has the (/)₈ TIM-barrel fold architecture typical of all plant endo-1,3--glucanases. The deep catalytic cleft is located in the upper part of the enzyme, over the C-terminal face of the barrel, and runs across the entire molecule. The side walls of the cleft are built up by the helices and loops. The most flexible region of the structure is the additional subdomain situated around helix 6 (Wojtkowiak et al., 2012). This subdomain, which forms part of the catalytic cleft, consists of loops located around helix 6 (6-6 and 5-5) and of additional secondary-structure elements located within these loops: two short antiparallel -strands located in the 5-5 loop and two short -helices located in loop 6-6.

3.3. Mass-spectrometric analysis of the hydrolytic products

The oligosaccharides located in the catalytic cleft of both complex structures are shorter than the hexasaccharide used for crystallization and must have arisen by hydrolysis. Therefore, hydrolysis was catalyzed by the protein despite the mutation of the active-site nucleophile. To confirm these observations, a mass-spectrometric (MS) analysis of the oligosaccharides was perfomed. The sample of the hexamer used in the crystallization trials contained predominantly hexasaccharide and a trace amount of pentasaccharide (data not shown). An MS analysis of oligosaccharides harvested directly from the crystals was unsuccessful, probably owing to the presence of polyethylene glycol or/and an insufficient amount of the target oligosaccharides. Therefore, a test of the hydrolytic activity of the E259A protein mutant was carried out in a separate experiment under conditions identical to the crystallization conditions except for the elimination of the polyethylene glycol component. The sample was analysed after two months of incubation, i.e. after a period of time equivalent to the duration of the crystallization experiment up to the point of data collection. As expected, the incubation mixture contained mainly tetrasaccharides (most abundant) and trisaccharides (Fig. 3). To confirm that the hexasaccharide does not undergo spontaneous hydrolysis, an identical test was performed without addition of the enzyme; the MS analysis of the incubation mixture excluded such a possibility (data not shown).

Figure 3 MALDI-TOF ionization mass-spectrometric analysis of the hydrolysis products of a hexasaccharide (13)--glucan catalyzed by the E259A mutant of GLUB20-2. M6, M5, M4 and M3, molecular masses of ionic adducts of Na+ and hexasaccharides, pentasaccharides, tetrasaccharides and trisaccharides, respectively; *, matrix ions (DHB, 2,5-dihydroxybenzoic acid; THAP, 2',4',6'-trihydroxyacetophenone).

3.4. Hydrolysis-product pattern and binding affinities

The endohydrolytic activity of the GLUB20-2 protein allows processing of the hexasaccharide substrate used in the crystallization experiments into two identical trisaccharide molecules or into a tetrasaccharide and a disaccharide. The production of trisaccharide molecules induced crystallization of the 3G complex, with the trimer bound at the nonreducing (glycon portion) end of the binding cleft. Potential complexes with a disaccharide bound at the glycon site were not formed, probably because in the incubation equilibrium the shorter product was outperformed in binding efficiency by the longer trisaccharide. Kinetic and thermodynamic studies of barley endo-1,3--glucanases (Hrmova et al., 1995) show an absence of any significant binding at positions beyond subsite -3, which is consistent with our results. With higher substrate/product concentration, the aglycon portion of the binding cleft is also occupied in crystal 7G. There is crystallographic evidence (disorder) suggesting that in this case the binding at subsites +1 and +2 is weak and that efficient docking is achieved only at the farther subsites (+3 and +4). There is a theoretical possibility that the +1, +2, +3 and +4 subsites in the 7G structure could be occupied by a disaccharide/trisaccharide/tetrasaccharide mixture with (combined) full occupancy only at subsites +3 and +4. However, since endo-glucanases are specific for the recognition and turnover of long saccharide chains, binding of a short oligomer far away (up to 10 Å) from the catalytic centre is not a likely scenario. It is obvious that the two product molecules modelled in the 7G complex could not have arisen in a single hydrolytic act, i.e. cleavage and product retention, of the hexasaccharide substrate. It must be thus assumed that the complex formation is a dynamic process that also involves diffusion (out and in) of the hydrolytic products, with the additional possibility of a transglycosylation reaction. Such an interpretation does not invalidate the crystallographic 7G model; on the contrary, it makes it even more useful because the oligosaccharides span seven binding subsites and not just six as would be possible with the original hexasaccharide. The oligosaccharide-enzyme interactions are described in detail in the following sections.

3.5. Protein interactions at the nonreducing end of the oligosaccharide

In both the 3G and 7G structures the trisaccharide molecule is situated in a negatively charged portion of the catalytic cleft and occupies subsites -1 to -3 (glycon portion). The electron density is very well defined, showing that all three glucose units adopt the relaxed chair conformation ⁴C₁ (Fig. 4). The glucose units at subsites -1 and -2 interact with the highly conserved residues Asn117, Glu310, Lys313 and Glu319. There are eight direct protein-ligand hydrogen bonds plus two water-mediated interactions for just these two subsites (Table 2). Asn117 is strictly conserved in clan GH-A enzymes, immediately preceding the catalytic proton donor Glu118 in the sequence. The importance of the amino-acid residues at sites Glu310, Lys313 and Glu319 for the hydrolysis reaction has been confirmed by site-directed mutagenesis studies of barley endo-1,3--glucanase (Chen et al., 1995). The presence of these residues is crucial for efficient catalysis. The enzyme-ligand interactions at the hydrophobic surfaces of the aromatic side chains are the characteristic ligand-recognition patterns found in many glycoside hydrolases. In the present case, the aromatic ring of Tyr58 makes contacts with glucosyl -2 and those of Phe305 and Phe322 make contact with the -1 glucose (Fig. 5). The position of Phe305 in the structure is defined by an unusual cis-peptide bond formed with the preceding Ala304. The interactions of hydroxyls O-1 and O-2 of glucose at subsite -1 with Wat259 were not taken into account because the presence of this water molecule is an artificial consequence of the side-chain reduction introduced by the E259A mutation, as described in the next section. The glucosyl residue at the -3 subsite forms three solvent-mediated and two direct interactions with the protein (Table 2). The O-3 and O-4 hydroxyls are directly hydrogen-bonded to the O¹ and N² atoms of Gln82, respectively. The tight binding and steric clashes around the O-3 hydroxyl group preclude the insertion of another glucosyl moiety at the -4 subsite, thus limiting the glycon product of the reaction to trisaccharides. This provides a plausible explanation of the preferences of this plant endo-1,3--glucanase for the trisaccharide products of the hydrolysis reaction. Gln82, which is part of a short flexible 3₁₀-helix preceding the third -helix of the TIM-barrel fold, is not conserved in other plant endo-1,3--glucanases. It is often replaced by serine or aspartate residues. It can therefore be expected that there will be differences between plant endo-1,3--glucanases in the binding mode of the glucosyl residue at subsite -3.

Table 2 Hydrogen-bonding and hydrophobic interactions with aromatic side chains between the oligosaccharide ligands and the protein Glucose unit at subsite Direct hydrogen bonds Water-mediated hydrogen bonds Hydrophobic interactions with aromatic rings -3 Gln82 O1O-3 Asn81 NO-2   Gln82 N2O-4 Lys32 NO-6     His61 N2O-6   -2 Asn117 N2O-2 Glu11 O1O-2 Tyr58-face# Lys313 NO-5 Lys32 NO-4   Lys313 NO-6     Glu310 O1O-6     -1 Lys313 NO-4   Phe305-face# Glu319 O1O-4   Phe322C-6 Glu319 O2O-6     Tyr201 OO-5     +3 Thr166 O1O-6 Tyr201 OO-2     Gly161 OO-4     Arg175 NHO-4   +4 Thr166 O1O-4 Gly205 OO-2   Thr166 OO-6 Ser240 OO-4   #In the conventional Haworth representation (McNaught & Wilkinson, 1997) of the glucose molecule, the -face is located beneath the pyranose-ring plane.

Figure 4 Atomic models of (13)--glucans. (a) The carbohydrate molecules bound in the catalytic cleft of the present 1,3--glucanase in the 7G complex. Electron density is shown as an Fo - Fc OMIT map (green) contoured at 2.5. The models in orange represent the glucose units built in electron-density maps and included in the coordinate set; the two units of the tetrasaccharide shown in grey (+1 and +2) are not part of the crystallographic model because of disorder and have been tentatively placed in the visible electron density in the binding groove only for the purpose of this figure (and Fig. 7) to illustrate the length and general shape of the ligand in subsites +1 to +4. (b) A heptamer modelled by Bohne et al. (1999) using SWEET2 (yellow) superposed on the nonreducing end of the trisaccharide of the present 7G complex.

Figure 5 The interactions of the trisaccharide ligand (orange) bound at the -1, -2 and -3 subsites of the protein. The key protein residues are shown in yellow and hydrogen bonds are marked by dashes. Note the site of the Ala259 mutation: the space freed by the side chain of Glu259 was occupied by the sugar unit -1, which is now in a relaxed 4C1 conformation.

3.6. The -1 subsite

The position of the glucose unit at subsite -1 is defined by interactions with strictly conserved residues (Table 2). The hydroxyl groups O-1 and O-2 interact only with Wat259. This situation is not typical for glucoside hydrolases and is a consequence of the mutation of the catalytic nucleophile Glu259 to alanine. The small side chain of the alanine residue allows the hydroxyls O-1 and O-2 of glucose -1 to slip into the empty space left by the glutamate (Fig. 5). A superposition of the complex and the native enzyme structure shows that hydroxyl O-1 would be at a distance of only 1 Å from the O² atom of Glu259. However, the interaction of the nucleophile O¹ atom with the O-2 hydroxyl of the substrate, which is strictly conserved in clan GH-A, is emulated in the 7G and 3G structures by Wat259, which is additionally hydrogen-bonded to the hydroxyl group of Tyr58 in the same manner as the Glu259 O¹ atom in the native structure. Another clan GH-A characteristic interaction of glucose residue -1 is a hydrogen bond between the O-2 hydroxyl and the N² atom of the asparagine residue preceding the proton donor (Asn117). In the complex structures of the mutated protein, the N² atom of Asn117 instead interacts with the O-2 hydroxyl of the glucosyl residue at the -2 subsite, but the position of the Asn117 side chain in the mutant and in the native protein is different.

The pyranoside rings in the -1 subsite undergo conformational changes during the hydrolysis reaction, as reviewed by Vocadlo & Davies (2008) and Davies et al. (2012). The pathway of the conformational changes typical for retaining -glycosidases is ¹S₃[⁴H₃]⁴C₁. To better analyze the interactions at the -1 subsite, the structures of the native and mutated GLUB20-2 protein were superposed and the -1 glucose conformation was changed from ⁴C₁ to ⁴H₃ (Fig. 6). With this new conformation docked in the native protein, the hydroxyls O-4, O-3 and the ring O atom O-5 establish exactly the same interactions with the enzyme as found for the chair conformation; in addition, the O-1 hydroxyl interacts with the O² atom of Glu259 and the O-2 hydroxyl interacts with Glu259 O¹ and Asn117 N², which is typical for the GH-A clan.

Figure 6 The interactions of the glucose units -2 and -1 with the strictly conserved GH-A clan residues, namely two glutamic acids, the catalytic nucleophile (259, here mutated to alanine) and the proton donor (118), and an asparagine (117) in front of the proton donor. (a) The structure of the present complex 7G shown in a 2Fo - Fc map (blue) contoured at 1. (b) The structure of the native protein (PDB entry 3ur7 ; grey) shown after superposition on the complex structure, with the glucose residue -1 modelled in the 4H3 conformation (grey). This altered conformation reduces the degree of steric conflict with the side chain of the Glu259 nucleophile, which coincides with a water molecule in the mutant structure.

3.7. Consequences of the E259A mutation

In the native structure of GLUB20-2 (PDB entry 3ur7 ) the O² (catalytic centre) and O¹ atoms of the catalytic nucleophile Glu259 form hydrogen bonds to Tyr201 O and to Tyr58 O, respectively. The proton donor Glu118 and the strictly conserved Asn117 form a hydrogen-bond network with Arg56 and Asn199 which stabilizes their positions in the structure, readying them for interactions with the incoming oligosaccharide (Fig. 7b). In the mutant complex structures 3G and 7G the side chain of Asn117 has moved away to form a hydrogen bond to the O-2 hydroxyl of the glucosyl residue at subsite -2, breaking part of the hydrogen-bond network mentioned above (Fig. 7a). However, as described previously, in the case of a properly placed glucosyl residue at subsite -1 the hydrogen-bond network will be preserved. The mutation of Glu259 to alanine has created sufficient space to accommodate a water molecule. This water molecule, Wat259, is hydrogen bonded to Tyr58, Arg56 and Asn199.

Figure 7 The interaction of the catalytic residues (a) in the 7G structure and (b) in the native GLUB20-2 structure (PDB entry 3ur7 ).

In many glycosidases, replacement of the catalytic nucleophile by a shorter residue allows small external anions such as azide, formate or halides to effectively cleave the glycosidic bond (Zechel & Withers, 2001). In the present mutant this is not the case since no modified oligosaccharides were detected in the mass-spectrometry experiments. We therefore postulate that the role of a nucleophile could possibly be played by a water molecule accommodated in a fashion similar to Wat259 but hydrogen-bonded to Tyr201 O, which would place it in the position of Glu259 O². This situation could lead to inversion of the configuration at the anomeric carbon; however, a second water molecule approaching from the opposite direction after the departure of the leaving group would invert the configuration again to the more stable form before diffusion of the final oligosaccharide product from the enzyme. A water molecule acting as a nucleophile has been observed, for example, in Micromonospora viridifaciens sialidase (Watson et al., 2003). There is also a slim possibility that the observed residual activity is the consequence of contamination by trace amounts of the native form introduced during protein purification or by amino-acid misincorporation during protein synthesis (Laughrea et al., 1987; Toth et al., 1988).

3.8. Leaving group-enzyme interactions

During the course of the hydrolytic reaction, the leaving group (aglycon portion, containing the reducing end of the original oligosaccharide substrate) diffuses away from the active centre and is replaced by a water molecule that is used to hydrolyze the glycosyl-enzyme intermediate (Fig. 2). In the 7G complex, the docking interactions anchoring the tetrasaccharide leaving group in the binding cleft are evidently weaker than at the complementary trisaccharide site. The glucosyl residues at the +1 and +2 subsites could not be modelled with reliability owing to their disorder (Fig. 3). Nevertheless, for glucosyl +1 a hydrophobic interaction with Phe204 can be expected. The glucose units at the +3 and +4 subsites form direct hydrogen bonds to a common residue, Thr166 (Table 2). Thr166 is located in the additional subdomain region of the 5-5 loop (Fig. 8), which is the most flexible part of the protein (Wojtkowiak et al., 2012). Our failure to crystallize a complex with a pentasaccharide may be explained by the insufficient length of the oligosaccharide, which was too short to establish efficient docking interactions; in consequence, the substrate would be unable to properly present the scissile bond to the catalytic centre.

Figure 8 The interactions of units +3 and +4 of the tetrameric (13)--glucan ligand (orange) with the protein. Hydrogen bonds are marked by dashed lines.

Interestingly, despite the weak interactions of the glucosyl residues at the +1 and +2 subsites of the binding cleft, the histidine tag in the crystal structures (PDB entry 3ur8 ; Wojtkowiak et al., 2012) of the native protein was bound there (Supplementary Fig. S1¹). The histidine-tag peptide spanned these two subsites and part of subsite -1, forming hydrogen bonds to the strictly conserved Tyr201 and Glu319 residues.

The kinetic studies reported for the barley enzyme indicated the presence of eight subsites, with the highest binding affinities at subsites -2, +4 and +5, where subsite +5 is located over the subdomain (Hrmova et al., 1995). The glucose residue at the +5 subsite was proposed to interact with the conserved Tyr167. In the present structures (as well as in the native protein models) the side chain of Tyr167 is exposed into the lumen of the catalytic cleft and is not hydrogen-bonded to any other residue, which endows it with high flexibility. It could possibly take part in the association with an extended oligosaccharide chain but, based on the structure and on the lack of evidence for binding of longer (pentameric and hexameric) saccharides there, we do not predict strong binding at this particular subsite.

3.9. Binding-cleft geometry defines substrate specificity of plant endo-1,3--glucanases

The long chains of natural (13)--glucans usually adopt a triple-stranded helical structure with 18 glucose units per pitch: six from each chain (Chuah et al., 1983). In this triple-helical structure, inter-chain hydrogen bonds are formed between three adjacent O-2 hydroxyl groups and intra-chain hydrogen bonds are formed between O-4 hydroxyls and ring O atoms O-5. For comparative studies, a model of a single (13)--glucan chain was prepared by Bohne et al. (1999) using the SWEET2 program for constructing three-dimensional models of saccharides from their sequences (Fig. 4b). Considered together, the trisaccharide and tetrasaccharide chains bound in the 7G complex form a more linear (less bent) structure, which is achived by a shortening of the O-4O-5 hydrogen bonds. As illustrated in Fig. 4(b), the oligosaccharide in solution modelled by SWEET2 has a distinct U-shape and cannot be superposed on the trace of the two oligosaccharide products present in the 7G complex. This suggests that the binding of a substrate molecule prior to the endo-1,3--glucanase reaction must involve a substantial deformation at the (-1)-(+1) linkage.

The geometry of the (13)- glycosidic bond defines the specificity of the present enzyme and of other plant endo-1,3--glucanases. For example, the cellulose chains with -1,4-glucosidic linkages are straight and the consecutive glucosyl residues are rotated by 180°, which results in the placement of the O-6 hydroxyls on both sides of the chains. In contrast, the O-6 hydroxyls of 1,3--glucans are located on the same side of the oligosaccharide chain. The shape of the substrate-binding cleft of the present protein is not compatible with binding of (14)--glucans, and -1,6 branching is only possible at the +1 and +2 subsites (Fig. 9).

Figure 9 Docking of (13)--glucans in the binding cleft of plant 1,3--glucanase. The protein, viewed from above the TIM barrel looking into the binding cleft, is shown as a green surface representation, except for the proton donor (118) and nucleophilic (259) residues, which are coloured red, and the subdomain region (magenta). The sugar units of the ligands that are included in the crystallographic model (7G) are shown in orange. The two units of the tetrasaccharide ligand that could not be modelled in electron density because of disorder are tentatively marked in this figure in grey (compare with Fig. 4).

The few crystal structures of longer (13)--oligosaccharides in protein complexes present in the PDB indicate that these oligosaccharides are preferentially recognized in a bent conformation similar to that proposed in solution (Supplementary Fig. S2). For example, the hexasaccharides found in carbohydrate-binding modules are described as `U-shaped' (Boraston et al., 2002; van Bueren et al., 2005) and the two pentasaccharides present in the catalytic cleft of Gas2p transglycosylase (Hurtado-Guerrero et al., 2009) also adopt such a conformation. The tetrasaccharide found in complex with laminaripentaose-producing -1,3-glucanase (Wu et al., 2009) was also described as `forming a distinct kink'.