J. Ševčík, Institute of Molecular Biology, Slovak Academy of Sciences, Dúbravská cesta 21, 84551 Bratislava, Slovakia Fax: +421 259307416 Tel: +421 259307435 E-mail: firstname.lastname@example.org
Most glucoamylases (α-1,4-d-glucan glucohydrolase, EC 126.96.36.199) have structures consisting of both a catalytic and a starch binding domain. The structure of a glucoamylase from Saccharomycopsis fibuligera HUT 7212 (Glu), determined a few years ago, consists of a single catalytic domain. The structure of this enzyme with the resolution extended to 1.1 Å and that of the enzyme–acarbose complex at 1.6 Å resolution are presented here. The structure at atomic resolution, besides its high accuracy, shows clearly the influence of cryo-cooling, which is manifested in shrinkage of the molecule and lowering the volume of the unit cell. In the structure of the complex, two acarbose molecules are bound, one at the active site and the second at a site remote from the active site, curved around Tyr464 which resembles the inhibitor molecule in the ‘sugar tongs’ surface binding site in the structure of barley α-amylase isozyme 1 complexed with a thiomalto-oligosaccharide. Based on the close similarity in sequence of glucoamylase Glu, which does not degrade raw starch, to that of glucoamylase (Glm) from S. fibuligera IFO 0111, a raw starch-degrading enzyme, it is reasonable to expect the presence of the remote starch binding site at structurally equivalent positions in both enzymes. We propose the role of this site is to fix the enzyme onto the surface of a starch granule while the active site degrades the polysaccharide. This hypothesis is verified here by the preparation of mutants of glucoamylases Glu and Glm.
In addition to catalyzing the removal of β-d-glucose from the nonreducing ends of starch and other related poly and oligosaccharides, glucoamylase is able to degrade α-1,6-glucosidic linkages, although much less effectively. The enzyme is produced by many moulds and yeasts. The primary industrial use of glucoamylase is in the production of glucose and fructose syrups, which in turn serve as a feedstock for biological fermentations in the production of ethanol or in the production of high fructose sweeteners . Using the classification of glycoside hydrolases into nearly 100 families on the basis of sequence similarity, glucoamylase belongs to family 15  (http://afmb.cnrs-mrs.fr/CAZY/).
The most thoroughly studied glucoamylase is that from Aspergillus awamori variety X100. The three-dimensional structure of its catalytic domain has been described in detail at a range of pH [3,4]. Subsequently, its interactions with different carbohydrate inhibitors were defined by the determination of structures in complex with 1-deoxinojirimycin , acarbose  and d-gluco-dihydroacarbose [7,8]. These structures define the positions of malto-oligosaccharide residues in at least the −1 and +1 subsites labeled according to the nomenclature proposed by  and identify interactions between substrates and active site amino acid side-chains. The structure of the starch binding domain of A. niger glucoamylase was solved by NMR in its native state  and in a complex with β-cyclodextrin . Crystal structures of an intact two-domain prokaryotic glucoamylase were determined from the clostridial species Thermoanaerobacterium thermosaccharolyticum with and without acarbose . In all of these enzymes the N-terminal starch binding domain has 18 antiparallel strands arranged in β-sheets of a super-β-sandwich, while the C-terminal catalytic domain is an (α/α)6 barrel.
Different strains of the dimorphous yeast Saccharomycopsis fibuligera produce a set of closely related glucoamylases. Two of them, (Glu; strain HUT7212) and Glm (strain IFO 0111) from the GLU and GLM genes, consist of 492 and 489 amino acid residues, respectively, with a sequence identity of 60% and a similarity of 77%, Fig. 1. The two enzymes differ in biochemical properties, in particular in the ability to digest raw starch. While Glu adsorbs to, but does not digest raw starch, Glm adsorbs well to starch granules and is capable of raw starch digestion. The glucoamylases from Aspergillus niger and A. awamori prefer longer malt-oligosaccharides as substrates, which is also the case for S. fibuligera glucoamylases .
The determination of the crystal structure of recombinant glucoamylase Glu at 1.7 Å resolution was reported earlier . The core of the enzyme is an (α/α)6 barrel known in SCOP nomenclature  as a six-helical hairpin toroid, and is closely similar to that of the catalytic domain of A. awamori and T. thermosaccharolyticum glucoamylases, with the active site at the narrower end of barrel. There is no terminal starch-binding domain, and this is clearly also true for the closely related Glm, for which a homology model was proposed . Thus the S. fibuligera glucoamylases Glu and Glm differ from the other characterized glucoamylases in that the raw-starch affinity site is an integral part of the single catalytic domain.
In this paper, two structures are described: that of the glucoamylase Glu with the resolution extended to 1.1 Å (Glu1.1) and that of its complex with acarbose at 1.6 Å resolution (Glu-A). One acarbose binds at the expected catalytic site, and we propose that the second site corresponds to the remote starch binding site. Five residues (Arg15, His447, Asp450, Thr462 and Tyr464) which are important in the remote starch binding site in Glu are conserved in Glm (Arg15, His444, Asp447, Thr459 and Phe461). However, a key residue which is central for the remote acarbose binding is different in the two enzymes: Tyr464 in Glu versus Phe461 in Glm (Fig. 1). To confirm that the remote binding site is essential for raw starch binding, the above amino acids were mutated and the mutants tested for their ability to adsorb to and digest raw starch.
Results and discussion
Description of the structures
There is one molecule in the asymmetric unit of both structures composed of a single domain consisting of 14 helices, 12 of them forming an (α/α6) barrel as expected from our previous native structure . The active site is at the narrower end of the barrel as mapped by the presence of ligands (Tris in Glu1.1 or acarbose in the Glu-A structure).
Accuracy of models
As expected, the accuracy of the structure Glu1.1 at atomic resolution is higher than that of Glu-A or Glu. The overall coordinate error for Glu1.1 and Glu-A estimated from the σA plot , estimated standard uncertainty (ESU) based on R and Rfree factors (the Cruickshank's dispersion precision indicator DPI , and the average temperature factors for protein atoms, water molecules and ligands are given in Table 1. The temperature factors are in good agreement with estimates from the Wilson plot .
Table 1. Refinement statistics. ESU, estimated standard uncertainty.
Molecules in asymmetric unit
Model – atom sites
Average B-values (Å2)
Wilson plot (Å2)
Coordinates ESU based on R/Rfree (Å)
σA error estimate (Å)
Stereochemical restraints r.m.s. (σ)
Bond distances (Å)
Bond angles (°)
Chiral centers (Å3)
Planar groups (Å)
Main-chain bond (Å2)
Main-chain angle (Å2)
Side-chain bond (Å2)
Side-chain angle (Å2)
The Ramachandran plot  calculated by the program procheck for Glu1.1 and Glu-A shows that in both structures, there are > 92% of residues in the most favored regions, the rest in additionally allowed regions except Ala339 and Ser357 which are in generously allowed regions. The electron density for both residues in the two structures is clear and all main-chain atoms are well ordered, which confirms that the deviation of torsion angles from ideal geometry of these two residues is an intrinsic feature of the structure. In Glu1.1 there is another residue, Ser305 in the generously allowed region. This residue is part of the loop Gly302–Ser306, which is poorly ordered in this structure (see below).
In both structures for most of the residues the ω angle deviates significantly from planarity. This is reflected in the G-factor calculated by procheck (Table 2) in which the ω angles score for Glu1.1 and Glu-A has a value of −0.05 and −0.06, respectively, with 489 contributors. This confirms that the peptide bond deviates from planarity by up to 20° as observed in a number of atomic resolution structures. The average value for ω angle in Glu1.1 and Glu-A structures is 179.6 and 179.5, respectively, with rmsd of 5.7° in both.
Table 2. G-factors calculated by procheck.
Dihedral angles (°)
Chi3 and Chi4
Main-chain covalent forces
Main-chain bond lengths (Å)
Main-chain bond angles (°)
The Glu1.1 structure
The Glu structure (1AYX) was described in detail previously. Superposition of the structures Glu1.1 and Glu based on all CA atoms, calculated by the program lsqkab, shows that the two structures are nearly identical with rmsd 0.38 Å. The maximum deviation (3.63 Å) does not represent any important difference as it relates to the C-terminal residue. Omitting 16 atoms from the surface loops for which deviation was above 1 Å, the rmsd falls to 0.32 Å. The superposition reveals that the molecule contracts on cryo-cooling with the surface regions being shifted towards the centre by ∼0.3 Å, keeping the central part of the molecule intact. This is reflected in the unit cell volume which is 510 156 Å3 at 292 K but falls to 479 022 Å3 at 110 K. Some of the residues poorly determined in the Glu structure became clearer in the Glu1.1 and all six residues with two conformations in Glu have a single conformation in Glu1.1.
Inspection of the Glu1.1 electron density shows that it is very clear in the entire molecule with only a single conformation for each residue suggesting that the molecule has a rigid fold. Nevertheless the segment Gly302-Glu303-Ser304-Ser305-Ser306 located at the opposite end of the barrel to the active site has weaker electron density and the temperature factors of the atoms in this segment are ∼31 Å2, 2.35 times above the average B for the structure. The high flexibility of this loop does not appear to be connected with the catalytic function. One explanation lies in the fact that the loop protrudes from the surface of the molecule and does not form any additional contacts with the molecule.
The Glu-A structure
The Glu-A structure was refined to a low R factor (Table 1) and the electron density is clear through the whole structure. While the Gly302-Ser306 loop has an average temperature factor of 26 Å2, compared with an average value for the whole protein of 13.7 Å2, the electron density is considerably better in comparison to the Glu1.1 structure. This is due to the close proximity of a phosphate anion (sodium phosphate buffer was used in purification) which fills the gap between the loop and the rest of the protein (Fig. 2B) forming a number of direct and water-mediated hydrogen bonds. One of the phosphate oxygen atoms forms hydrogen bonds with a water molecule which belongs to the cluster of water molecules and the Asp379 carboxyl liganded to a Na+ ion. Another Na+ ion, surrounded by five water molecules is bound to Ile177 carbonyl. All distances between the Na+ and the surrounding oxygen ligands are close to 2.42 Å, the average distance observed in a set of protein structures .
Superposition of the Glu-A and Glu1.1 structures gives r.m.s. and maximum displacement of 0.47 and 4.47 Å, respectively. Glu-A differs from Glu1.1 mainly in the loop Ser9-Asn10-Tyr11-Lys12-Val13-Asp14-Arg15-Thr16 where the differences between CA atoms are up to 4.5 Å (at Asn10). This conformational change is caused by Arg15 which moves (CA moves 1.2 Å) in order to interact with the acarbose sugar +1 causing reorientation of the whole loop.
The catalytic reaction of glucoamylases proceeds with inversion of configuration at the anomeric carbon which requires a pair of carboxylic acids at the active site, one acting as general acid and the other as general base . The mechanism of hydrolysis consisting of three steps involves proton transfer to the glycosidic oxygen of the scissile bond from a general acid catalyst, formation of oxocarbenium ion and a water-assisted nucleophilic attack by a general base catalyst [24–27]. In the glucoamylase from A. awamori and A. niger Glu179 was identified as the general acid and Glu400 as the general base [4–6,28,29]. Superposition of the A. avamori and A. niger structures with those of S. fibuligera glucoamylase complexes with Tris and acarbose shows that the corresponding residues are Glu210, general acid and Glu456, general base. In the Glu-A and Glu1.1 structures the distances between the CA atoms of these two residues are 14.8 and 14.7 Å and the shortest distances between the two carboxyl groups are 7.3 and 7.6 Å, respectively. The carboxyl groups can easily adopt a distance of 9.2 Å, typical for inverting glycoside hydrolysis [24,30,31].
In the active site of the native Glu1.1 there is a Tris molecule which forms direct hydrogen bonds with Arg69, Asp70 and one bond, mediated by a water molecule, with Glu210. Hydrogen bonds formed between the enzyme and Tris are the same as observed previously .
In the Glu-A complex there are two acarbose molecules: one in the active site and the other on the surface of the enzyme about 25 Å away, Fig. 2. The active site acarbose fits tightly into the pocket (Fig. 3) and the electron density for all the acarbose atoms is very clear (Fig. 4A). The acarbose has a well-defined conformation that corresponds to that observed in the complex with the fungal glucoamylase from A. awamori var. X100 at pH 4 . The sugars −1 and +1, labeled according to the nomenclature proposed by , form several hydrogen bonds with the enzyme and confirm the identity of the active site residues. Sugars +2 and +3 do not form any hydrogen bonds with the enzyme, however, they do stack nicely against the aromatic rings of Tyr351 and Trp139, respectively. The distances between the sugars and the aromatic rings of the two residues are ∼4 Å. The mode of acarbose binding to the active site readily explains the exoglucanase activity.
Raw starch binding site
The electron density for the surface acarbose (Fig. 4B), is not as clear as that for the active site acarbose, suggesting a higher mobility or a reduced occupancy, probably caused by a neighboring molecule at a distance of about 3.5 Å. This is reflected in the average temperature factors which are 33 Å2 for the surface acarbose in contrast to the 13 Å2 for the active site ligand. The surface acarbose in the Glu-A structure, which we propose to correspond to a raw starch binding site, is localized in the crevice formed by Arg15, His447, Asp450, Thr462, Tyr464 and Ser465. There are six H-bonds between this remote acarbose and the enzyme, two direct, His447 ND1 – O3 (+ 2), Thr462 O – O2 (+ 2) and four mediated through one or two water molecules, Asp450 N–W – O3 (+ 1), Asp450 OD1–W – O2 (−1), Asn451 N–W–W-O4 (−1), Ser465 N–W – O3 (+ 1). The second sugar ring of the acarbose stacks against the planar Arg15 guanidino group.
A space-filling model of glucoamylase with both acarbose molecules is shown in Fig. 5. The surface acarbose is curved around Tyr464 in the form of a semicircle (Fig. 6) and captures the inhibitor molecule as seen in the ‘sugar tongs’ binding site in barley α-amylase isozyme 1 complexed with the substrate analogue, methyl 4′,4′′,4′′′-trithiomaltotetraoside [32,33] and a true oligosaccharide substrate . A similar situation was seen in the structure of the amylomaltase–acarbose complex [35,36] in which the acarbose molecule winds around Tyr54. However, in those structures the raw starch binding site is not part of the catalytic but is located on a separate domain.
Mutations at the remote ligand binding site
To verify the hypothesis that the site on the Glu surface interacting with acarbose represents the starch binding site, the point mutants R15A, H447A, T462A and a double mutant H447A, D450A were prepared and tested for affinity to starch. Two approaches were used: adsorption of enzymes in a test tube assay on a native granular starch and mobility of enzymes in native gels with and without copolymerized boiled granular starch.
Adsorption of the wild-type Glu, its mutants and Glm in test tube experiments is presented in Fig. 7. The results show that affinity of Glu to native raw starch was observed only at a high raw starch–enzyme ratio: at a ratio of 100 mg raw starch−50 µg Glu only 10% of the enzyme was bound. Under the same conditions, > 95% of the wild-type Glm was bound. The Glu mutants did not bind at all.
The electrophoretic mobility of Glu and its mutants are presented in Fig. 8. In a standard native gel (Fig. 8A) the Glu and its mutants move to nearly the same position while in the gel with a copolymerized boiled granular starch (Fig. 8B) all Glu mutants move significantly faster indicating that their affinity to the gel matrix is lower.
As documented in our previous work , a similar situation was found with the raw starch degrading Glm. The Glm H444A, D447A mutant in the gel containing starch moved faster than wild-type Glm because of its impaired affinity towards the substrate. The changes in electrophoretic mobility of native and mutant glucoamylases demonstrate that mutations of the amino acids proposed to be involved in binding of the surface acarbose caused reduction of enzyme adsorption on starch, proving that these amino acids are involved in starch binding site in spite differing in a key residue – Tyr464 in Glu versus Phe461 in Glm. Biochemical analysis has shown that the double mutation H444A, D447A retained specific activity on soluble starch identical but caused significant reduction of raw starch hydrolysis (to 12%) in comparison with the wild-type enzyme.
The structures of the glucoamylases from S. fibuligera belong to family 15 of the glycoside hydrolases. Most of the currently characterized family members have a two-domain structure, the small domain playing the role of binding the enzyme to starch, allowing the larger catalytic domain to hydrolyze the starch substrate. We showed previously that the S. fibuligera Glu enzyme lacked the independent starch binding domain while the catalytic domain was very similar to that of other family 15 members. The close similarity in sequence of the Glm enzyme indicated that it too lacked the binding domain, and the modeled structure was like that of Glu with a single domain.
Our present work has improved the resolution of the native Glu structure, but has in addition revealed the presence of a second acarbose (substrate analogue) binding site on the surface of the enzyme, 25 Å remote from the catalytic site. The key residues involved in the binding at this remote site have been mutated, and the mutants shown to have greatly reduced starch binding properties. These results strongly support the hypothesis that the S. fibuligera glucoamylases have evolved a starch binding site on the catalytic domain quite distinct from that seen in other family 15 glycoside hydrolases.
In vitro mutagenesis
Site-directed mutagenesis was performed by QuickChangeTM site-directed mutagenesis kit (Stratagene, La Jolla, USA). Plasmid pVT100L-Glu  was used as a template. The following oligonucleotides were used: GLU R15A forward (5′-ATTCAAACTATAAAGTTGACGCAACTGACTTGGAAACCTTC-3′), GLU R15A reverse (5′-GAAGGTTTCCAAGTCAGTTGCGTCAACTTTATAGTTTGAAT-3′); GLU H447A forward (5′- GCAAGTCATTTTGGATGCTATTAATGATGATGGCTC-3′), GLU H447A reverse (5′- GAGCCATCATCATTAATAGCATCCAAAATGACTTGC-3′); GLU T462A forward (5′- GAACAACTTAACAGATATGCCGGTTATTCCACCGGTGCC-3′), GLU T462A reverse (5′- GGCACCGGTGGAATAACCGGCATATCTGTTAAGTTGTTC-3′); GLU H447A, D450A forward (5′-GCAAGTCATTTTGGATGCTATTAATGCTGATGGCTCCTTGAATGAAC-3′), GLU H447A, D450A reverse (5′-GTTCATTCAAGGAGCCATCAGCATTAATAGCATCCAAAATGACTTGC-3′).
All mutations were verified by DNA sequencing.
Enzyme preparation and purification
The recombinant glycosylated glucoamylases were prepared in Saccharomyces cerevisiae AH22 as described previously [14,38]. Yeast transformants were grown in medium containing 1% yeast extract, 2% peptone, 2% glucose, for 48 h. Proteins which showed electrophoretic homogeneity were obtained from extracellular media after ultrafiltration through Amicon PM-30 membrane, molecular sieving chromatography on Superose 12P and ion exchange chromatography on FQ (both from Amersham Bioscience, Vienna, Austria).
Polyacrylamide gel electrophoresis
Polyacrylamide gel electrophoresis was performed under native conditions. Concentration gel was omitted. Two types of gels were used: (1) Standard 10% polyacrylamide gel: 1.25 mL of 1.5 m TrisHCl buffer, pH 8.8, 1.45 mL of water, 2.2 mL of acrylamide solution (30%), 60 µL of 10% ammonium persulfate solution and 2.5 µL N,N,N′,N′-tetramethylethylendiamine (TEMED) were mixed together. (2) Polyacrylamide gel (7.5%) with copolymerized boiled granular corn starch: a suspension of 37.5 mg of starch in 1.25 mL of 1.5 m TrisHCl buffer, pH 8.8, and 2 mL of water was boiled for 5 min and after cooling to room temperature, 1.65 mL of acrylamide solution (30%), 60 µL of 10% ammonium persulfate solution and 2.5 µL TEMED were added. The positions of glucoamylases were detected with Coomassie Brilliant Blue R-250 staining (Merck, Darmstadt, Germany).
Raw starch binding assay
The purified enzymes, in amounts of 0.01–0.5 mg mL−1 protein, were added to a suspension of 100 mg of raw corn starch in 1 mL of 0.05 m sodium acetate at pH 5.6 and 4.5 for Glu and Glm, respectively, which are optimal values for soluble starch hydrolysis. The mixture was gently stirred for 1 h at +4 °C. After centrifugation at 13 000 g for 5 min, the protein content expressed as enzyme activity of the supernatant was assayed. The amount of the bound protein was calculated from the difference between the initial enzyme activity and the free enzyme activity in the supernatant after binding.
Glucoamylase activity was determined in the reaction mixture containing 0.9% Leulier soluble starch in 0.05 m sodium acetate, pH 5.6 and 4.5 for Glu and Glm, respectively, incubated with enzyme at 40 °C for 15 min. An increment of glucose was measured as described previously .
Crystallization, data collection and processing
The recombinant nonglycosylated Glu was prepared essentially as reported in . The enzyme was crystallized from a protein solution of 10 mg·mL−1 in 50 mm acetate buffer at pH 5.4 and 15% PEG 8K, as described earlier .
Protein for preparation of the glucoamylase–acarbose complex was isolated in the same way as before, but Tris was replaced by sodium phosphate buffer to avoid Tris binding at the active site. Native crystals of the enzyme were prepared as above and then 1 µL of the mother liquor enriched by acarbose at a concentration of 10 mm was added to drops (5 µL) containing native crystals a few days before data collection.
X-ray data from native and complex crystals were collected at 110 K on EMBL beam lines BW7B to 1.1 Å and X11–1.6 Å resolution, respectively, at the DORIS storage ring (DESY, Hamburg, Germany). Each data set was collected from a single crystal with a MAR Research (Hamburg, Germany) imaging plate scanner and processed with denzo and scalepack. A summary of data collection and processing is given in Table 3.
Table 3. Data statistics. Values in parentheses refer to the highest resolution shell.
All subsequent calculations were performed with programs from the CCP4 package  unless otherwise indicated. As the unit cell parameters of glucoamylase at 1.1 Å resolution (Glu1.1) and the glucoamylase–acarbose complex (Glu-A) were slightly different from those of Glu (1AYX), molecular replacement molrep, was used to position the model in the new cells. Both structures were refined with the program refmac against 95% of the data with the remaining 5% randomly excluded for cross-validation using the free R factor (Rfree) . All data were included in the final refinement step. After each refinement step, ARP  was used for modeling and updating the solvent structure.
The Glu1.1 and Glu-A structures were initially refined with isotropic temperature factors and in the later stages with anisotropic temperature factors including the contributions from the hydrogen atoms. Hydrogen atoms were generated according to established geometrical criteria on their parent C, N and O atoms. The temperature factors of the hydrogen atoms were set equal to those of their parent atom. Isotropic and anisotropic temperature factors, bond lengths, and bond angles were restrained according to the standard criteria employed by refmac. Occupancies of water molecules were set to unity and not refined. The models were adjusted manually between refinement cycles on the basis of (3Fo-2Fc, αc) and (Fo–Fc, αc) maps using the programs o and xtalview. The refinement statistics are given in Table 1.
Modeling of the structure
A model of the glucoamylase Glm structure was generated using the modeller w4 package  using the known structure of glucoamylase Glu and the sequence similarity between the two enzymes .
Data Bank accession numbers
The atomic coordinates have been deposited in the Protein Data Bank for Glu-A (2F6D) and Glu1.1 (2FBA). GenBank accession no(s) M17355 and AJ311587 belong to GLU and GLM genes, respectively.
This work was supported by Howard Hughes Medical Institute grant no. 75195–574601 and the grants 1/0101/03 and 2/1010/96 awarded by the Slovak Grant Agency VEGA.