Oligosaccharide Binding to Barley α-Amylase 1*

Enzymatic subsite mapping earlier predicted 10 binding subsites in the active site substrate binding cleft of barley α-amylase isozymes. The three-dimensional structures of the oligosaccharide complexes with barley α-amylase isozyme 1 (AMY1) described here give for the first time a thorough insight into the substrate binding by describing residues defining 9 subsites, namely -7 through +2. These structures support that the pseudotetrasaccharide inhibitor acarbose is hydrolyzed by the active enzymes. Moreover, sugar binding was observed to the starch granule-binding site previously determined in barley α-amylase isozyme 2 (AMY2), and the sugar binding modes are compared between the two isozymes. The “sugar tongs” surface binding site discovered in the AMY1-thio-DP4 complex is confirmed in the present work. A site that putatively serves as an entrance for the substrate to the active site was proposed at the glycone part of the binding cleft, and the crystal structures of the catalytic nucleophile mutant (AMY1D180A) complexed with acarbose and maltoheptaose, respectively, suggest an additional role for the nucleophile in the stabilization of the Michaelis complex. Furthermore, probable roles are outlined for the surface binding sites. Our data support a model in which the two surface sites in AMY1 can interact with amylose chains in their naturally folded form. Because of the specificities of these two sites, they may locate/orient the enzyme in order to facilitate access to the active site for polysaccharide chains. Moreover, the sugar tongs surface site could also perform the unraveling of amylose chains, with the aid of Tyr-380 acting as “molecular tweezers.”

Structural studies of ␣-amylases in complex with substrate analogues have received much attention in the past decade (1)(2)(3)(4)(5)(6)(7)(8)(9)(10)(11)(12)(13). The pseudotetrasaccharide inhibitor acarbose was used for the vast majority of these studies. In contrast, binding of natural substrates was only described in three complexes known to date. These are the ␣-amylase from porcine pancreas complexed with maltopentaose (14,15), the inactive catalytic mutant E208Q ␣-amylase from Bacillus subtilis in complex with maltopentaose (16), and the "maltogenic" ␣-amylase from Bacillus stearo-thermophilus in complex with a maltose unit that was derived from a maltotriose (17). ␣-Amylases belong to the glycoside hydrolase family 13 (afmb.cnrs-mrs.fr/CAZY), which together with GH70 and GH77 constitute GH clan H (18). These enzymes share the catalytic site geometry with three invariant acid residues as follows: a catalytic nucleophile, a proton donor, and a third catalytic acidic side chain, which polarizes the glucoside unit at subsite Ϫ1 (19). Despite a sequence identity of 80% and a nearly identical overall folding (20), AMY1 and AMY2 exhibit important differences as concerns stability (21,22), enzymatic properties (23), and sensitivity to the proteinaceous inhibitor barley ␣-amylase subtilisin inhibitor, belonging to the Kunitz soybean trypsin inhibitor family (24). Most interestingly, three calcium ions are bound to both isozymes sharing identical ligands (20), although calcium affects the activity of AMY1 3 and AMY2 in distinct manners (21). Numerous AMY1 subsite mutants have been characterized with respect to substrate affinity and catalytic capacity, but structural insight for barley ␣-amylases was limited until now to subsites Ϫ1 through ϩ2 being experimentally defined with the AMY2-acarbose complex (4). Enzymatic subsite mapping (25) and computer-aided modeling (26,27) suggested that both AMY1 and AMY2 possess 10 subsites spanning from the nonreducing end at subsite Ϫ6 to ϩ4 on the aglycon accommodating part. A more recent study in which crystals of AMY1 soaked in a solution containing thio-maltotetraose was expected to expand our knowledge on subsite-binding modes in the catalytic cleft of plant ␣-amylases. This substrate analogue, however, led to the discovery of a new surface binding site at domain C (see Fig. 1), the so-called "pair of sugar tongs," but did not bind to the active site (28). This site and an earlier discovered surface binding site on the catalytic domain made up of two consecutive tryptophan residues, Trp-278 and Trp-279 (AMY1 numbering), perform distinct interactions with the sugar rings of different oligosaccharide ligands (4,28). Mutational analyses have confirmed that these sites indeed can bind onto starch granules and that they also bind ␤-cyclodextrin (29,62). Here we report the crystal structures of native AMY1 as well as an inactive mutant of the catalytic nucleophile Asp-180 in complex with acarbose (Fig. 2, A and B) and maltoheptaose ( Fig. 1), respectively. This is the first report on sugar binding to the active site of barley ␣-amylase 1 and on binding of the substrate maltoheptaose to the active site region.

EXPERIMENTAL PROCEDURES
Preparation of Recombinant AMY1⌬9-A truncated form, AMY1⌬9 (nonapeptide deletion from the C terminus), of recombinant AMY1 was prepared as described previously (30) to overcome difficulties encountered in growing three-dimensional crystals of full-length AMY1 (30). AMY1⌬9 is henceforth referred to as AMY1.
Production and Purification of AMY1 D180A -The selected transformant was grown (0.5 liters of BMGY, 2 days in a 5-liter flask) to A 600 ϭ 20, and the medium was replaced by BMMY (1 liter) for induction for 32 h under vigorous shaking. The supernatant was kept, and the cells were resuspended in BMMY (1 liter) for a second induction culture (39 h). AMY1 D180A was purified from the combined supernatants by affinity chromatography on ␤-cyclodextrin-Sepharose (31, 32) and anion FIGURE 1. Stereo drawing of the overall structure of AMY1 D180A inactive mutant in complex with maltoheptaose. Calcium ions are indicated as green spheres. Catalytic residues are highlighted in pink. The full maltoheptaose molecule bound in the catalytic cleft is represented by a gray transparent surface. To the left, the starch granule-binding surface site (occupied by a maltopentaose molecule) is shown with the two tryptophan residues (Trp-278 and Trp-279) highlighted in blue. In domain C (bottom part of the figure), another maltopentaose molecule is curved around Tyr-380 (in blue) at the level of the sugar tongs surface site. FIGURE 2. A, atoms labeling convention for glucosyl residues and acarbose. The acarviosine unit is constituted by rings A and B, including the amino group valienamine, and are ␣-1,4-linked to rings C and D representing a maltose unit. B, superimposition of residues implicated in the substrate binding in active sites of AMY1 and AMY2 (stereo view). Only residues interacting directly by hydrogen bonds are shown. The complex AMY2acarbose (4) (Protein Data Bank entry 1BG9) is presented in blue, AMY1-acarbose (this work, Protein Data Bank entry 1RPK) in red, AMY1-thio-DP4 (28) (Protein Data Bank entry 1P6W) in green, and AMY1 D180A -acarbose (this work, Protein Data Bank entry 1RP9) in yellow. exchange chromatography (32). The enzyme was concentrated (Centriprep YM10, Millipore, Bedford, MA) to ϳ2 mg ϫ ml Ϫ1 and stored in 10 mM MES, 5 mM CaCl 2 , pH 6.8, 0.02% (w/v) sodium azide at 4°C. Isoelectric focusing (pI 4 -6.5) and SDS-PAGE (PhastGels and Phast-System, Amersham Biosciences) of AMY1 D180A were silver-stained according to the manufacturer's recommendation (36). Enzyme concentrations were calculated from amino acid contents of protein (25 g) hydrolysate (32). Purified AMY1 D180A was obtained (6.5 mg ϫ liter Ϫ1 ) and gave one band in SDS-PAGE (45 kDa) and native PAGE (data not shown).
Crystallization, Soaking, and Cryo-protection-The complex between AMY1 and the inhibitor acarbose was obtained by co-crystallization at conditions derived from those of the native enzyme (30). 0.5 l of a 100 mM acarbose stock solution was mixed with 2 l of protein stock and 3 l of well solution, thus resulting in a final concentration of 9.1 mM acarbose in the drop, and crystals grew to a final size of 0.35 ϫ 0.1 ϫ 0.05 mm 3 within 3 weeks.
The inactive AMY1 D180A mutant was crystallized by the hanging drop vapor diffusion method at 19°C, using protein stock (2.9 mg ϫ ml Ϫ1 ) in 10 mM MES, pH 6.7, 5 mM CaCl 2 , and 0.02% NaN 3 . Drops were prepared by mixing 2 l of protein stock with 0.5 l of 3% (v/v) isopropyl alcohol and 2.5 l of well solution containing 20% (w/v) polyethylene glycol 8000. Thin crystals typically grew to a final size of 0.8 ϫ 0.1 ϫ 0.05 mm 3 after ϳ6 months. Soaking was done by adding a solution of maltoheptaose (Roche Applied Science) directly to the drop to a final concentration of 10 mM and leaving it for 24 h at 19°C.
The AMY1 D180A -acarbose complex was obtained by adding acarbose directly to the drop at a final concentration of 10 mM and leaving it for 24 h at 19°C. All crystals were cryo-protected prior to data collection by rapid soaking in three successive steps in mother liquor containing increasing concentrations of ethylene glycol (5, 10, and 15% (v/v)) and 10 mM acarbose or maltoheptaose, respectively.
Diffraction Data Collection-AMY1-acarbose diffraction data were collected on a MARresearch 345 Image Plate System associated to a Nonius FR591 rotating anode (CuK ␣ radiation), operating at 44 kV and 100 mA and coupled to Osmic confocal mirrors. Data processing and reduction was carried out using DENZO (38) and SCALA from the CCP4 package (39). AMY1 D180A -maltoheptaose data were collected at the ID14-1 beamline at the European Synchrotron Radiation Facility, Grenoble, France, on a MarCCD detector, and data on the AMY1 D180A -acarbose were collected at the FIP BM30A beamline (European Synchrotron Radiation Facility) on a MarCCD detector. Diffracted intensities were integrated with the program MOSFLM (40) as implemented in the CCP4 software package (39) and scaled with SCALA (39). All crystals belong to the orthorhombic space group P2 1 2 1 2, and one molecule is present in the asymmetric unit. Data collection statistics are presented in TABLE ONE.
Structure Determination and Refinement-Because of crystal isomorphism with AMY1 (28), the latter was used as starting model (Protein Data Bank entry 1HT6) in a difference Fourier, where all water molecules, calcium ions, and ligand molecules had been removed. For all data sets, an initial rigid body refinement, including data to 4 Å resolution, was performed, and in the remaining refinements a simulated annealing protocol was used extending the data to 2.0 Å resolution. All refinements were done with the program CNS (41). In order to avoid over-refinement, free and conventional R-factors were monitored (42). Alternating with these refinement steps, visual examination of electron density maps and manual building was carried out using the graphic software TURBO-FRODO (43). Based on the inspection of 2F o Ϫ F c and F o Ϫ F c maps (contoured at 1 and 3, respectively), calcium ions were inserted, and water molecules were manually added respecting hydrogen bonds with standard distances and angles formed to appropriate atoms. Ligands (sugar moieties from acarbose or maltoheptaose) were manually inserted in the electron density maps. Because of with k, a scaling factor. R free is calculated from a test set constituted by 10% of the total number of reflections randomly selected.
the high quality and resolution of the diffraction data, the orientation and nature of each sugar ring was unequivocally determined. Models of acarbose and glucose units were found in the HIC-Up data base (44). The geometry and coordinate errors of the three final structures were examined with the programs PROCHECK (45) and WHATCHECK (46). Refinement statistics are listed in TABLE ONE. Structural superimposition of the structures was performed using the "rigid" option in TURBO-FRODO (43). Figures shown in this paper were rendered using ISIS-Draw (freeware from Elsevier MDL, San Leandro, CA; www.mdli.com), TURBO-FRODO (43), and VIEWER-LITE TM 4.2 (freeware from Accelrys Inc., San Diego; www.accelrys.com).

AMY1 in Complex with Acarbose and Comparison with AMY2-Acarbose-
The three-dimensional structure of the complex between AMY1 and acarbose was solved to 2.0 Å resolution. As for AMY2 (4), three substrate binding subsites are occupied by a molecule of component 2 (a pseudotrisaccharide derived from acarbose after cleavage of a glucose unit from the reducing end) meaning that unit A to C occupies subsites Ϫ1 to ϩ2. The starch granule-binding surface site containing Trp-278 and Trp-279 also binds a component 2 molecule, whereas the "sugar tongs" site at domain C (28) displays the entire acarbose molecule. Direct hydrogen bonds between enzyme and inhibitor molecules are listed for the three sites in TABLE TWO. The electron density map clearly suggests that acarbose was cleaved by AMY1 resulting in a component 2 molecule binding to the active site. At the starch granule-binding surface site, the fourth sugar ring is not defined, reflecting either that this sugar partly points into the solvent and therefore is flexible or a difference in affinity for this and the sugar tongs-binding site, respectively. This conclusion is supported by the structures, the surface site with Trp-278 to Trp-279 being widely opened and exposed to solvent. Only a relatively low number of interactions are present for stabilizing long flexible substrates; therefore, a short and less flexible substrate may have a higher affinity. In contrast, the sugar tongs site is defined by a cleft in which oligosaccharide is captured by Tyr-380 and has an increased number of interactions compared with the former surface site, thus a priori providing superior stabilization for longer ligands. Remarkably, when comparing the active sites for AMY1-acarbose and AMY2-acarbose, structural conservation between the enzymes is high as only two amino acid side chains differ between the isozymes (see Fig. 2B). Arg-183 AMY1 replaces Lys-182 AMY2 and Asn-209 AMY1 corresponds to Ser-208 AMY2 , without major consequence for substrate binding, because all side chains form hydrogen Barley ␣-Amylase Oligosaccharide Complexes SEPTEMBER 23, 2005 • VOLUME 280 • NUMBER 38 bonds to acarbose. Some variation, however, is seen in the conformation of the oligosaccharides as demonstrated by the torsion angles, ⌽ and ⌿, describing the flexibility around the ␣-(1,4)-glucosidic linkage. ⌽ between the cyclitol (ring A) and the 4,6-dideoxy-␣-D-glucose (ring B) units in acarbose is defined by C6-C1-N4Ј-C4Ј and ⌿ is defined by C1-N4Ј-C4Ј-C5Ј, whereas between 4,6-dideoxy-␣-D-glucose (ring B) and ␣-D-glucose (ring C) units, these are ⌽ ϭ O5-C1-O4Ј-C4Ј and ⌿ ϭ C1-O4Ј-C4Ј-C5Ј. In AMY1-acarbose, these values are (⌽,⌿) ϭ (18°,Ϫ142°), ring A(Ϫ1) to ring B(ϩ1), and (⌽,⌿) ϭ (97°,Ϫ120°), ring B(ϩ1) to ring C(ϩ2), compared with AMY2-acarbose with (⌽,⌿) ϭ (Ϫ2°,Ϫ122°), ring A(Ϫ1) to ring B(ϩ1), and (⌽,⌿) ϭ (90°,Ϫ114°), ring B(ϩ1) to ring C(ϩ2). The ⌽,⌿ values found in the AMY1-acarbose complex are in good agreement with those reported for other ␣-amylase-acarbose complexes around the glucosidic bond corresponding to the scissile bond in substrates (47), with (⌽,⌿) ϭ (18°,Ϫ148°), in the complex between the psychrophilic Pseudoalteromonas haloplanktis ␣-amylase and the transglycosylation product derived from acarbose (8), and (⌽,⌿) ϭ (32°,Ϫ151°) for a homologue complex with Thermoactinomyces vulgaris R-47 ␣-amylase 2 (48). The difference in ⌽,⌿ values found in the AMY2-acarbose complex may be related to the lower resolution of the crystal structure (2.8 Å) as compared with the others. AMY1 D180A in Complex with Acarbose-The purified inactive mutant AMY1 D180A showed a specific activity of 6.7 ϫ 10 Ϫ4 mol ϫ min Ϫ1 ϫ mg Ϫ1 , corresponding to 6.7 ϫ 10 Ϫ4 % of wild-type AMY1 having a specific activity of 100 mol ϫ min Ϫ1 ϫ mg Ϫ1 . In the complex between acarbose and AMY1 D180A , with Asp-180 being the catalytic nucleophile, an intact acarbose molecule occupies the three sugar binding sites as follows: the active site, the starch granule-binding surface site, and the sugar tongs. The high quality of the electron density maps allows unambiguous definition and orientation of all sugar units. In the active site, acarbose (Fig. 2B) covers subsites Ϫ4 to Ϫ1. Eleven direct hydrogen bonding interactions exist between AMY1 D180A and the inhibitor, eight of these being in subsite Ϫ1 (ring D), one to ring C (subsite Ϫ2), and finally two to ring A (subsite Ϫ4). The conformation of the tetrasaccharide differs from analogue sugar moieties in the acarbose-derived trisaccharide in the AMY1-acarbose complex. ⌽,⌿ values for the acarbose chain are (107°,Ϫ116°), ring A(Ϫ4) to ring B(Ϫ3); (70°,Ϫ152°), ring B(Ϫ3) to ring C(Ϫ2); and (83°,Ϫ153°), ring C(Ϫ2) to ring D(Ϫ1). These ⌽,⌿ values clearly demonstrate that the acarbose sugar chain adopts dissimilar conformations depending on the subsite to which it is bound. An important difference that probably is a consequence of the mutation of the catalytic nucleophile is the loss of hydrogen bonding between His-93 in subsite Ϫ1 and the aca-O6A atom as found in the complex AMY1-acarbose or an analogue sugar moiety occupying this subsite (see TABLES TWO, part A, and THREE, part A). His-93 also lacks this hydrogen bond in the maltoheptaose complex described below. This suggests that the nucleophile is essential for the stabilization of the Michaelis complex to obtain the "V"-shaped orientation of glucose molecules residing in subsites Ϫ1 and ϩ1. The nucleophile furthermore plays a key role in the stabilization of the abovementioned conserved His-93 for which the imidazole ring must adopt a particular conformation in order to hydrogen bond to the substrate, and thereby stabilize the complex. Earlier mutational analysis indeed indicated that this histidine, which belongs to a conserved motif in GH13 (49), is important for transition state stabilization (29).
AMY1 D180A in Complex with Maltoheptaose-The three-dimensional structure of inactive AMY1 D180A -maltoheptaose exhibits the same three distinct sugar-binding sites as the complexes with acarbose (see Figs. 1, 3, and 4). In the active site, an entire molecule of maltoheptaose resides at subsites Ϫ1 to Ϫ7. Both the starch granule-binding surface site and the sugar tongs, however, only show five of the seven glucose units.
In the active site the orientation of the glucose units (Glc2000 to Glc2006, from the nonreducing toward the reducing end) is readily interpretable. The overall conformation of maltoheptaose can be considered as a twisted "S" (see Fig. 3A). All glucose units at subsites Ϫ1 to Ϫ7 adopt a chair conformation. The enzyme kinetics subsite maps did not identify subsite Ϫ7 that was proposed in a study reporting 10 subsites from Ϫ7 to ϩ3 (50). For Glc2000 (in subsite Ϫ7) to Glc2003 (subsite Ϫ4), the hydroxyl groups at C6 point into the bulk solvent, whereas those of Glc2004 (subsite Ϫ3) to Glc2006 (subsite Ϫ1) point toward the interior of the protein. The conformation of the heptasaccharide as described by ⌽,⌿ values are (118°,Ϫ113°) Ϫ Glc2006(Ϫ7) to Glc2005(Ϫ6); (123°,Ϫ111°) Ϫ Glc2005(Ϫ6) to Glc2004(Ϫ5); (Ϫ62°,Ϫ73°) Ϫ Glc2004(Ϫ5) to Glc2003(Ϫ4); (122°,Ϫ122°) Ϫ Glc2003(Ϫ4) to Glc2002(Ϫ3); (69°,Ϫ148°) Ϫ Glc2002(Ϫ3) to Glc2001(Ϫ2); (83°,Ϫ155°) Ϫ Glc2001(Ϫ2) to Glc2000(Ϫ1). When compared with the analogue values in AMY1 D180A -acarbose, ⌽ and ⌿ are rather similar between subsites Ϫ3 to Ϫ1 as expected. However, when approaching the nonreducing end of the glycone part of the active site, only ⌽,⌿ values between subsites Ϫ2 to Ϫ1 are similar to the analogue ones as found in other complexes (47). Hereafter they differ, which is particularly remarkable when comparing to (⌽,⌿) for the interglycosidic bond between subsites Ϫ3 to Ϫ2 in the B. subtilis ␣-amylase-maltopentaose structure (16) being (123°,Ϫ107°). This seems to be mainly due to the torsion angles between subsite Ϫ5 to Ϫ4 glucoses, being subjected to a very drastic shift resulting in (⌽,⌿) ϭ (Ϫ62°,Ϫ73°) deviating to a very high degree from those observed in a regular helical structure of amylose. In this structure the repeating unit consists of a maltotriose where the conformation of the glycosidic linkage is (91.8°,Ϫ153.2°), (85.7°,Ϫ145.3°), and (91.8°,Ϫ151.3°) (51). The interactions between maltoheptaose and active site residues are listed in TABLE THREE, part A. Eight of the 16 direct hydrogen bonds between maltoheptaose and AMY1 D180A involve the reducing end ring (Glc2006), whereas Glc2004 has no direct interactions with the enzyme, and an aromatic stacking exists between Trp-10 and Glc2005 (subsite Ϫ2). In comparison to this relatively low number of contacts, AMY1-acarbose (see TABLE TWO, part A) has 15 direct hydrogen bonds between the enzyme and the three observed rings of acarbose, as well as a certain number of interactions mediated by water molecules. Fig. 4 shows a schematic representation, summarizing these interactions. Furthermore, a number of internal hydrogen bonds are observed for maltoheptaose in the complex (see TABLE THREE, part D). A summary of residues defining subsites Ϫ1 to Ϫ7 is given in TABLE FOUR. The fact that no direct interactions exist between sugar moieties in subsite Ϫ3 (Glc2004) and the enzyme is in excellent agreement with the binding kinetics of malto-oligosaccharides, which allows determination of subsite affinities as listed in TABLE FOUR. Two independent groups have arrived at somewhat different results (25,50), although some features are shared. These studies show that the affinity at subsite Ϫ3 is low compared with the other subsites (25) and even negative (50). An unexplained density around Cys-95 in the native structure of AMY1 (28) is seen in the electron density here as well but does not seem to affect the interaction with Glc2002.
The presence of subsite Ϫ7 in the AMY1 D180A -maltoheptaose complex was not unequivocally defined by subsite mapping (25,50). In this subsite, His-45 and Val-47 make direct hydrogen bonds to the nonreducing end of maltoheptaose (Glc2000), see Most interestingly, Glc2001 is located between two hydrophobic residues Val-47 and Tyr-105, but no hydrophobic contacts exist to either of these. The space between these two residues may be considered as the substrate "entrance" to the active site cleft. The plane of ring Glc2001 is not parallel to that of the aromatic group of Tyr-105 and an imperfect aromatic stacking is present between these two entities. Probably, this stacking would be optimized if the additional glucose unit (Glc2000) at Barley ␣-Amylase Oligosaccharide Complexes SEPTEMBER 23, 2005 • VOLUME 280 • NUMBER 38 Direct hydrogen bond contacts in the complexes AMY1 D180A -maltoheptaose and AMY1 D180A -acarbose Atom labeling convention for glucosyl residues and acarbose is presented in Fig. 2A the nonreducing end of Glc2001 in maltoheptaose was lacking. It seems that Glc2000 forces Glc2001 to move away from Tyr-105 and thereby weakens the aromatic stacking interaction. Comparative studies of active site residues in AMY1 D180A -maltoheptaose and native AMY1 show that the backbones are superimposable and that no drastic reorientation of side chains occurred. All residues involved in interactions with the substrate and their neighbors are highly superimposable with the exception of Arg-183. In native AMY1, this residue displays a double conformation, for which the side chain points in two opposite directions. In the AMY1 D180A -maltoheptaose complex, Arg-183 has only one conformation with the side chain pointing toward Glc2006. Although Arg-183 does not interact directly with Glc2006, it contributes to the formation of a network of water molecules that mediates contacts between the above-mentioned glucose unit and neighboring residues. The side chain of Lys-130 is slightly reoriented compared with native AMY1 and makes a water-mediated hydrogen bond (Wat-896) to Glc2003. Finally, a very small shift is observed for Tyr-105, which approaches Glc2001 (see above). Therefore, it appears that substrate binding to AMY1 does not induce a significant reorganization of the protein residues.
When comparing the maltoheptaose complex to AMY1 D180A -acarbose, all atoms from rings C and D (the maltose unit in acarbose) are perfectly superimposed, in agreement with their respective ⌽,⌿ values. Moreover, all contacts established in the complex with maltoheptaose are invariant in the acarbose complex (see TABLE THREE, part A) for rings C and D, and the hydrogen bond lengths are very similar. Accordingly, backbones and side chains of both complexes are perfectly superimposable at subsites Ϫ1 and Ϫ2. Noticeably, the previously discussed Arg-183 is orientated in an opposite direction in the complex AMY1 D180A -maltoheptaose. This reorientation results in a slight modification of the water molecule network surrounding ring D in acarbose, without any major influence on the indirect water-mediated proteininhibitor interactions. As it was shown in the AMY1 D180A -maltoheptaose complex, acarbose ring B, which corresponds to glucose unit Glc2004, occupies subsite Ϫ3 and has no interaction at all with the protein. Once again, these two sugar rings are perfectly superimposed. Finally, ring A from acarbose is located in subsite Ϫ4 and its conformation slightly differs from Glc2003, its counterpart in AMY1 D180A -maltoheptaose, thus leading to an additional interaction with Ala-146-O. This difference is due to the inter-cyclic nitrogen atom in the acarviosine unit. Moreover, the presence of extra glucose units in maltoheptaose occupying subsites Ϫ5 to Ϫ7 seems to force the substrate to adopt a half-circle conformation centered on Val-47 (see above), which is not the case for acarbose. This is consistent with residues Val-47 and Tyr-105 defining the entrance of the catalytic pathway, thus serving as "guides" for the substrate and conferring the spatial organization for its approach toward the active site. Amino acid residues defining subsites in the active site cleft of AMY1 based on the structural studies of AMY1 D180A /maltoheptaose, AMY1 D180Aacarbose, and AMY1-acarbose complexes Catalytic residues are underlined. Binding kinetics in the various subsites as obtained by two independent groups (25,50)   Starch Granule Surface-binding Site-A secondary binding site at the surface of the enzyme was visualized earlier in crystal structures of AMY2 (4) and AMY1 (28) in complex with the substrate analogues acarbose and thio-maltotetraose, respectively. Two adjacent tryptophans, 278 and 279 (Trp-276 and Trp-277 in AMY2), define this socalled starch granule-binding site and were identified by differential labeling and UV spectroscopy (52). It has been shown that these two tryptophan residues are spatially locked because of their environment. An angle of 135°found between the two planes defined by the aromatic tryptophan side chains appears to be constant in both AMY1 and AMY2, in complexed as well as in native states (28). It was therefore suggested that Trp-278 and Trp-279 (AMY1) function as a kind of "geometric filter" determining whether substrates should bind to the active site.
The complex between the inactive mutant AMY1 D180A and maltoheptaose displays electron density corresponding to five of the seven glucose units (Glc4000 to Glc4004, numbered from the nonreducing to the reducing end, Fig. 3B). For this sugar molecule which forms a halfcircle, no electron density was observed for the two remaining glucose rings, probably because of their location in the bulk solvent, and thus suggesting a highly disordered state for these two rings. Gln-227 and Asp-234 are performing two major interactions as shown (TABLE  THREE, part B). Furthermore, stacking interactions are present between Trp-278/Glc4003 and Trp-279/Glc4002, respectively, and a network of water molecules mediates several indirect contacts between AMY1 and maltoheptaose. Comparative studies show that the only differences between native and complexed AMY1 at this site concern a small reorientation of the side chain of Gln-227 (to interact with Glc4003), which in turn gives rise to a small reorientation of Asn-231.
In the complex between the inactive mutant and acarbose, the starch granule-binding surface site also accommodates an intact acarbose molecule with rings B and C stacking onto Trp-279 and Trp-278, respectively. These interactions are in excellent agreement with those defined in AMY1 D180A -maltoheptaose (see TABLE THREE, part B), where rings B and C perfectly superimpose with maltoheptaose glucose units Glc4002 and Glc4003, but where the spatial location is less conserved for acarbose rings A and D, leading to a less curved conformation of the molecule as compared with maltoheptaose. Consequently, an additional interaction is found between Lys-271 and acarbose (see TABLE THREE, part B).
The Pair of Sugar Tongs-The third substrate-binding site is the AMY1-specific sugar tongs located in domain C (28). At this site, identified in a complex between AMY1 and a thio-maltotetraose substrate analogue, thio-DP4, Tyr-380 shows a shift upon sugar binding, resulting in a movement of the short loop defined by residues surrounding Tyr-380. This substrate binding ability is lacking in the known structures of AMY2, and in part may stem from Pro-376 AMY2 , which replaces Ser-378 AMY1 and thus rigidifies the loop and impedes its movement. Moreover, comparative studies of the tertiary structure of domain C in several ␣-amylases (28) showed that this domain C-binding site is unique to AMY1-type plant ␣-amylases. In fact, the presence of two additional ␤-strands in ␣-amylases from other species blocks the putative access for the substrate at this site. Moreover, Tyr-380, one of the key residues in the sugar tongs site, is conserved only in plant ␣-amylases (28).
In the AMY1 D180A -maltoheptaose complex, electron density is seen for five of the seven glucose units (labeled Glc3000 to Glc3004, numbering from the nonreducing end to the reducing end), see Fig. 3C. As in AMY1-acarbose, the sugar molecule has the shape of a half-circle with the Tyr-380 side chain pointing into the center. This malto-oligosaccharide is exposed into the solvent, which explains the lack of electron density for two sugar moieties. Direct hydrogen bond interactions are listed in TABLE THREE, part C along with intramolecular hydrogen  bonds stabilizing the molecule (TABLE THREE, part D). Moreover, several indirect interactions mediated by water molecules are present (results not shown). The binding of the substrate in this site causes small rearrangements as indicated by shifts of 0.8 and 3 Å for Tyr-380-C␣ and Tyr-380-OH, respectively. Finally, Thr-392 reorients its side chain to interact with Glc3001.
When comparing to the AMY1 D180A -acarbose complex, an entire acarbose molecule superimposes quite well with maltoheptaose on glucose units Glc3001, 3002, 3003, and 3004, respectively, with a large number of conserved interactions for the inactive mutant complexes at this site (see TABLE THREE, part C). Ring A in acarbose is not perfectly superimposed with Glc3001 because of the different chemical structure of the valienamine compared with a glucose ring. Ring D is rotated 180°c ompared with Glc3004, resulting in its hydroxyl group at C6 pointing in the opposite direction of its homologue in Glc3004. Remarkably, the locations of backbone and side chains, including Tyr-380, are fully conserved between the two complexes. Compared with the native structure of AMY1, the shift observed for Tyr-380-C␣ and Tyr-380-OH atoms in the acarbose complex is 1.2 and 3.4 Å, respectively, which is in excellent agreement with those observed in AMY1 D180A -maltoheptaose, as well as with those reported for AMY1-thio-DP4 being 1.2 and 3.1 Å, respectively (28).

DISCUSSION
The combination of structural studies of enzyme-substrate and enzyme-inhibitor complexes as revealed by the AMY1 D180A -maltoheptaose, AMY1 D180A -acarbose, and AMY1-acarbose complexes leads to a structural definition of subsites Ϫ7 to ϩ2, as summarized in TABLE FOUR. The present study reveals subsite Ϫ7, which was not unambiguously determined neither by substrate mapping (25) nor by molecular modeling (26,27). Accordingly, only subsites ϩ3 and eventually subsite ϩ4 remain to be visualized experimentally by a complex in AMY1 or AMY2. Our results are predominantly consistent with earlier observations from computer-aided modeling studies showing the interaction of a maltodecaose in the active site cleft of AMY2 (26) and AMY1 (27). These modeling studies proposed a hypothetical fork shape of the nonreducing end of the binding area, which should allow the substrate to reach the catalytic site by two distinct ways, or accommodation of ␣-1,6-branched substrates. Neither is confirmed by the present structures because no electron density corresponding to sugar rings is observed in the putative alternative binding region. The subsite mapping studies show low affinity at subsite Ϫ7, but biochemical data can prove the existence of a functional Ϫ7 subsite as small amounts of p-nitrophenyl and substantial amounts of glucose are released from PNPG7 (25) and malto-octaose (50), respectively. AMY1 has similar K m values for G7 and G8 (1.9-fold higher for G7) despite the 18-fold lower k cat /K m for G7 compared with G8 (50), which may be explained by nonproductive binding of G7 to subsites Ϫ7 to Ϫ1 as illustrated here in the AMY1 D180A -maltoheptaose complex. Binding modes of G7 and acarbose covering subsites Ϫ7 to Ϫ1 and Ϫ4 to Ϫ1, respectively, in AMY1 D180A are unexpected, however, from the subsite mapping. The affinity at subsite ϩ2 is obviously higher than at subsites Ϫ3 and Ϫ4 (25,50), suggesting acarbose binding to occur in subsites Ϫ2 to ϩ2, or otherwise that the nonreducing end valienamine ring that mimics the distorted glucose should be accommodated in subsite Ϫ1 as found for the wild-type AMY1-acarbose complex. The actual binding modes therefore must be due to the loss of the nucleophile. In both complexes the glucose moiety at the reducing end is located in subsite Ϫ1, is undistorted, and lacks the hydrogen bond between O6 and His-93, as observed in all complexes of the ␣-amylase family enzymes including the wild-type AMY1-acarbose complex. Asp-180 therefore seems to be essential both for distortion of the glucose ring in subsite Ϫ1 and for binding this distorted glucose ring in the appropriate orientation, resulting in substrate binding in a productive mode spanning subsites Ϫ1 and ϩ1. In AMY1 D180A , the loss of Asp-180 hence causes lack of distortion and thereby from binding in the productive binding mode.
AMY1 is more efficient than AMY2 in degrading starch granules and has higher affinity for substrates in general. Structural analysis of individual subsites within the substrate binding crevice of the two isozymes does not provide an explanation for this variation. However, the presence of the sugar tongs-binding site in AMY1, so far not observed in AMY2, makes a fundamental difference. This study demonstrates the ability of the sugar tongs surface site to recognize and bind natural substrates, whereas the previous study only showed binding of a substrate analogue (28). It also confirms binding of acarbose or maltoheptaose to the starch granule surface-binding site. The orientation of maltoheptaose molecules on the two surface substrate-binding sites in AMY1 suggests that both of them a priori are independent of each other. First, they are separated by a considerable distance, and it is difficult to imagine a single polysaccharide chain connecting these sites. Second, the reducing ends of oligosaccharides located in these two sites are facing each other if we try to build a link between these sites by the shortest path. The same conclusion is made for connecting molecules at the active site and the surface sites. This is only possible if the polysaccharide chain coils up around the enzyme, which seems highly improbable. AMY1 thus could interact with three distinct sugar chains, as opposed to AMY2 binding only two sugar chains.
The identification of two surface binding sites in AMY1 leads to the question of their role in vivo. Hypothetically, these two sites may allow interaction of the enzyme with amylose and amylopectin molecules in starch. We have shown that the substrates have a salient tendency to circularize themselves when binding to these sites, and if trying to complete this curved sugar chain to obtain a full cycle, it can be shown that ␤-cyclodextrin (7 glucose rings) and most probably ␣-cyclodextrin (6 glucose rings) are obtained at the sugar tongs-binding site. At the starch granule-binding site, ␤-cyclodextrin seems the more probable, in accordance with earlier studies demonstrating that AMY2 can bind ␤-cyclodextrin (52,53). This, however, needs experimental confirmation. The structure of amylose is a helicoidally arranged chain, each turn containing 6 glucosyl residues (54). It therefore seems plausible that the two surface sites in AMY1 can interact with amylose in its natural conformation.
Because of the specificity of these two sites, they may locate/orient the enzyme in order to facilitate access to the active site for polysaccharide chains. In addition, the sugar tongs surface site could also disentangle polysaccharide chains, Tyr-380 acting as "molecular tweezers" by its insertion in the helical and/or lamellar structure of starchy substrates (55).
These conclusions are supported by comparative studies with other family 13 glycoside hydrolases. For example, the structures of T. vulgaris R-47 ␣-amylase 1 (TVAI) complexed with malto-oligosaccharides reveal the presence of a domain "N" putatively acting as a starch-binding domain (56). As compared with the ␣-amylase domains A-C, this extra N-terminal domain N was shown to bind malto-oligosaccharides at two distinct sites, site N and site NA. The first one could interact with the outer surface of the starch helix, mainly through stacking interactions to aromatic residues, whereas the second one holds the saccharide units from both outside and inside of the helix by stacking interactions and hydrogen bonds. These authors suggest that site N is implicated in recognizing the surfaces of rigid helical structure of starch, whereas site NA may recognize the loose helical structure region or contribute to unravel helical starch. Also, specific hydrolytic activity of TVAI compared with that of ␣-amylase from Aspergillus oryzae (lacking domain N) is around 18-fold higher on raw starch, supporting the crucial role of domain N (56). The architecture of sites N and NA exhibit no similarity to that of the sugar tongs site in AMY1, because the substrate is not captured by a flexible aromatic residue entering its inner curvature. These sites resemble more closely the starch granule-binding surface site in AMY1 and AMY2, notably by the implication of at least one tryptophan residue performing aromatic stacking interactions with substrate. We suggest that the sugar tongs from AMY1 and site NA from TVAI may share a common role, whereas the starch granulebinding surface site in AMY1/AMY2 and site N could have a similar function.
Structural comparative studies were furthermore performed to cyclomaltodextrin glucanotransferases (CGTase, EC 2.4. 1.19), also belonging to the glycoside hydrolase family 13, and having, in addition to the property of hydrolyzing ␣-D-(1,4)-glycosidic bonds, the ability of circularizing oligosaccharides into ␣-, ␤-, or ␥-cyclodextrins. The crystal structure of the double mutant E257Q/D229N of CGTase from Bacillus circulans (strain 251) in complex with ␥-cyclodextrins showed that Tyr-195 in the active site is essential for the circularization mechanism (57). Remarkably, the phenolic ring of this residue is very close to the center of the ␥-cyclodextrin, and the binding of the molecule induces an important shift (2.6 Å) of this key residue. The similarity with the AMY1 sugar tongs, however, ends here, as Tyr-195 from the CGTase only makes hydrogen bonds with ␥-cyclodextrin but no hydrophobic interactions. CGTases bind ␣or ␤-cyclodextrins on two distinct surface sites located in domain E (58, 59) of the carbohydratebinding module family 20 (18). Two adjacent tryptophans (Trp-616 and Trp-662) in the CGTase perform aromatic stacking onto two glucosyl units, constituting one of the binding sites of this domain (site 1). Tyr-663, which makes hydrophobic interactions, and Leu-600, which inserts in the cyclodextrin cylinder (59), define the second site. The structure of binding site 1 shares the two tryptophanyl residues with the starch granule surface-binding sites in AMY1 and AMY2, but the CGTase has a very different environment, and the two tryptophans are less stabilized by neighboring residues. The side chains of these two residues possess a high degree of flexibility, allowing the accommodation of both ␣-, ␤-, and ␥-cyclodextrins in contrast to AMY1 and AMY2 counterparts, which are in a perfectly locked position (20,28). Evolutionary aspects may explain these common structural features between this CGTase and AMY1, as CGTases derived from ␣-amylases, keeping their hydrolytic activity and gaining their circularization property by the addition of new domains (60).
C domains of AMY1 and the glucosyltransferase amylosucrase from Neisseria polysaccharea have been compared. Amylosucrase bind substrates to its domain C (61), but with a distinct binding mode and location of the substrate as compared with AMY1. The ␤-sandwich domain C of amylosucrase does not confer enough space for accommodating the substrate, and the polysaccharide chain is bound onto the side of the domain, implicating among others a hydrophobic interaction with a phenylalanine residue.
Finally, binding modes in a complex between acarbose and amylomaltase from Thermus aquaticus (47) were compared with those reported herein. Amylomaltase is a member of GH 77, which together with GH 13 is a part of the clan H. It catalyzes either the transglycosylation with transfer from one ␣-1,4-glucan to another or an intramolecular cyclization resulting in much larger cyclodextrins than produced by CGTases. Two acarbose molecules are bound in this structure (47), one in the active site and a second close to the active center at a distance of 14 Å. In this latter site, key interactions determining the conformation and bind-ing of the inhibitor are the hydrophobic contacts of Tyr-54 and Tyr-101 with unit C from acarbose. The authors propose that Tyr-54 may help in curving the glucan chain, thus favoring synthesis of cyclic products. When leaving the catalytic center, the chain could "wrap around" Tyr-54 before returning to the active site, which appeared consistent with the formation of the smallest cycloamyloses having less than 22 glucan units. Recognition, binding, and circularization schemes of acarbose around Tyr-54 are very close to those observed at the sugar tongs site in AMY1. Because this region of amylomaltase cannot be considered as a separate domain, but as an extension of the active site, the similarities end here. Moreover, the role of this region seems to be circularization of the glucan chain, rather than recognizing such a conformation as we suggest for the AMY1 sugar tongs.
Major conclusions drawn from this comparative study are that recognition and binding modes of both surface sites in AMY1 are common to family members. These binding sites do not have their own catalytic properties but appear to contribute to enhance the activity of the enzyme. AMY1 seems to be the simplest enzyme in terms of the threedimensional structure that contains two distinct surface binding sites indirectly implicated in the catalytic activity. As an example, CGTase from B. circulans has also two surface sites, but they are located in the extra domain E and are not present in AMY1. Thus, AMY1 appears to be a highly "optimized" enzyme with an excellent compromise between catalytic efficiency on both soluble and insoluble starch-related substrates and structural complexity, because of the presence of the newly discovered sugar tongs-binding site.