Automated Docking of -(1,4)- and -(1,6)-Linked Glucosyl Trisaccharides in the Glucoamylase Active Site

To further characterize the molecular basis of glucoamylase selectivity, low-energy conformers of five -(1,4)- and -(1,6)-linked glucosyl trisaccharides were flexibly docked into the glucoamylase active site using AutoDock 2.2. To assure that all significant conformational space was searched, the starting trisaccharide conformers for docking were all possible combinations of the corresponding disaccharide low-energy conformers, as such starting structures have produced disaccharide-glucoamylase complexes comparable to those obtained by protein crystallography. All docked trisaccharides occupied the first and second subsites in very similar modes to those of corresponding nonreducing-end disaccharides. For linear substrates, full binding at the third subsite ccurred only when the substrate reducing end was -(1,4)-linked, hydrogen-bonding with the hydroxymethyl group being the only polar interaction there. For the one docked branched substrate, steric hindrance in the -(1,6)-glycosidic oxygen suggests that the active-site residues have to change position for hydrolysis to occur. The second subsite of glucoamylase active site allows flexibility in binding, but, at least in Aspergillus glucoamylases, the third selectively binds substrates -(1,4)-linked between the second and third subsites.

Glucoamylase [-(1,4)-D-glucan glucohydrolase, EC 3.2.1.3, GA] is industrially used to digest liquefied starch, an important step in the production of high-fructose corn syrup. GA releases -D-glucose by hydrolyzing terminal -(1,4)-D-glucosidic bonds from the nonreducing ends of maltooligosaccharides. Aspergillus niger GA can cleave -(1,6)-D-glucosidic bonds, which initiate amylopectin branches, at lower rates, besides hydolysing other -linked-glucosidic bonds (Meagher and Reilly, 1989). At high D-glucose concentrations GA forms differently linked condensation products (Nikolov et al., 1989), including the -(1,4)- and -(1,6)-linked trisaccharides maltotriose, panose, and isomaltotriose. -(1,4)-Linked products are kinetically favored, while that of products containing -(1,6)-linkages is preferred thermodynamically (Nikolov et al., 1989). Understanding the molecular basis of GA interaction with its many possible substrates and products is necessary to effectively engineer GA selectivity.

The crystal structures of complexes of the active site of Aspergillus awamori var. X100 GA with the inhibitors acarbose and D-gluco-dihydroacarbose (Aleshin et al., 1994, 1996; Stoffer et al., 1995) revealed the locations of the first four subsites in their interaction with maltooligosaccharides, with a pH-dependent dual conformation found beyond the second subsite. The interactions of methyl -maltoside and methyl -isomaltoside in the first and second subsites have been elucidated (Lemieux et al., 1996; Coutinho et al., 1997b,c). However, the indication that the residues at the third and fourth subsites are significantly less conserved (Coutinho and Reilly, unpublished results) demands a better understanding of the interaction of these subsites with different substrates.

Automated docking with AutoDock (Goodsell and Olson, 1990; Goodsell et al., 1993, 1996) of of D-glucose, D-mannose, and D-galactose and some monosaccharide-like inhibitors in GA (Coutinho et al., 1997a), yielded binding modes close to those found by X-ray crystallography. A combination of conformational analysis with docking to model the interaction of disaccharide substrates and their analogues with GA 1) yielded binding modes of methyl -acarviosinide (Coutinho et al., 1997a) that matched the disaccharide moiety of the inhibitor acarbose in its GA complex; 2) gave consistent binding modes of different isomaltosyl analogues, identifying the key hydroxyl group in the reducing end of isomaltose essential for correct binding in the second subsite (Coutinho et al., 1997b); and 3) correctly identified the real substrates among all - and b-linked glucosyl disaccharides, confirming that a key substrate hydroxyl group bound at the second subsite correctly positions the glycosidic oxygen for hydrolysis and is essential for enzyme action (Coutinho et al., 1997c).

Conformational analysis confirmed that the two-bond-linked maltose is much less flexible than the three-bond-linked isomaltose (Dowd et al., 1992, 1994). This study extends to trisaccharides the systematic study of the structural interaction of substrates with GA, clarifying the role of the third subsite in substrate binding and catalysis.

Conformers representing all possible combinations of the low-energy minima described in conformational studies of -maltose (3) and methyl -isomaltoside (7) were optimized by MM3(92). The number of conformers of each trisaccharide varied (see Figure 1): methyl -maltotrioside (9), methyl -panoside (21), methyl -isopanoside (21), methyl -isomaltotrioside (49), and methyl -(6¹--glucopyranosyl)-maltoside (21).

The torsional angles representing ring orientations about the -(1,4)-glucosidic linkage are defined as  ij = (H-1i-C-1i-O-4j-C-4j) and  ij = (C-1i-O-4j-C-4j-H-4j), while those about the -(1,6)-glucosidic linkage are ij = (H-1i-C-1i-O-6j-C-6j), ij = (C-1i-O-6j-C-6j-C-5j), and  ij = (O-6j-C-6j-C-5j-H-5j) (Figure 1), with i denoting the residue toward the nonreducing end and j denoting the one toward the reducing end. The pairs ij will be designated AB, BC, or A'B, for the pairs of glucosyl residues docked in A. awamori var. X100 GA active site at the first and second subsites, second and third subsites, and first branching subsite and second subsite, respectively.

Automated docking was performed using AutoDock 2.2 (Goodsell and Olson, 1990; Goodsell et al., 1993, 1996) (Scripps Research Institute, La Jolla, Cal.) in the GA crystal structure from the complex with inhibitor D-gluco-dihydroacarbose (Stoffer et al., 1995; Aleshin et al., 1996) (Brookhaven Protein Databank entry 1gai), with all water molecules excepting the putative catalytic molecule removed, in a 30-Å cubic box using a two-stage approach (Coutinho et al., 1997a,b,c). In the first stage, however, the nonreducing-end pyranosyl ring of each initial conformer was fixed. All structures of each trisaccharide were subjected to cluster analysis with a tolerance of 1 Å for an all-atom root mean square (RMS) deviation.

For the second stage, or redocking, the global minimum structure and the low-energy structures of the significant clusters were subjected to docking under similar conditions, but now permitting limited translation, followed again by cluster analysis. As before, substrate internal energies in both docking and redocking were referenced to the MM3(92) relative energy. Since both nonreducing ends of methyl -(6¹--glucopyranosyl)-maltoside can potentially fit in the first subsite of the GA active site, both cases were independently docked there.

The clusters obtained during redocking with total energies of interaction of the representative structure within 5 kcal/mol of the best interacting structure of each trisaccharide are listed in Table 1. Following analysis of the representative clusters, the structures believed to represent significant binding modes in the GA active site are depicted in Figure 2.

The docked trisaccharides bind often in modes similar to those of their corresponding nonreducing-end disaccharide moieties (Coutinho et al., 1997a,b). For methyl -panoside only the third cluster gives values in the same range, in what will be considered a productive binding mode. While automated docking allows only a productive class of binding modes at the first and second subsites for long substrates with nonreducing-end terminal -(1,4) linkages, there is significant variability in those substrates with -(1,6)-linked nonreducing-end residues, with clearly unproductive modes caused by contributions from interactions taking place beyond the second subsite.

For productive cases, the binding modes beyond the second subsite depend on the type of glycosidic bond. For linear trisaccharides, the two-bond linkage of the -(1,4)-bound disaccharide moiety at the reducing end allows a binding mode at the third subsite close to two slightly different conformations found there in crystallographic studies of the GA complex with D-gluco-dihydroacarbose (Stoffer et al., 1995; Aleshin et al., 1996). Such is the case for the second cluster of methyl -maltotrioside and the third cluster of methyl -panoside (Figures 2a and 2b). There we find, as in the GA complexes with acarbose and D-gluco-dihydroacarbose (Aleshin et al., 1994, 1996; Stoffer et al., 1995), hydrogen bonding between the exocyclical hydroxyl group 6C-OH and the backbone of both Gly121 and Glu179. Given that this is the only polar point of attachment, and that the hydroxymethyl group containing 6C-OH is flexible by nature, the dual conformation found there for acarbose and D-gluco-dihydroacarbose is not surprising.

The values of BC/BC obtained for the second docked cluster of methyl -maltotrioside are close to the values found at pH 4.0 for the most abundant conformers of acarbose and D-gluco-dihydroacarbose, respectively (Stoffer et al., 1995; Aleshin et al., 1996). For methyl -panoside, even though it is not possible to make a direct comparison of dihedral angles, it is very significant that the third residue of the top productive structure takes a position intermediate to that found for both conformers of D-gluco-dihydroacarbose at the third subsite. The GA active site can accommodate methyl -maltotrioside and methyl -panoside almost equally well, although a significant difference in distance exists between the two glycosyl units at the nonreducing end.

For methyl -isopanoside and methyl -isomaltotrioside, the longer -(1,6)-glycosidic bond in the reducing-end disaccharide moiety causes the results to differ, leading to a variety of possible binding modes (Table 1). Different hydrogen bonding can be found for these compounds between one or more of the hydroxyl groups at the third residue and the backbone of Gly121 or Glu179. Typical cases are the top structures of methyl -isopanoside and methyl -isomaltotriose, (Figures 2b and 2c), but there are alternative hydroxyl groups for different clusters. The three-bond linkage allows greater conformational liberty than the third subsite in the A. awamori var. X100 GA active site is designed to accommodate.

In the top productive methyl -(61--glucopyranosyl)-maltoside (Figure 2d), the branching -(1,6)-linked glycosyl unit has a mildly constrained conformation.

In the first two subsites, the interactions are very similar to those described earlier for docked methyl -maltoside and methyl -isomaltoside (Coutinho et al., 1997b,c). Polar interactions are stronger at the set of critical OH-4A, OH-6A, and OH-3B groups for the -(1,4)-linked nonreducing-end disaccharide moiety of methyl -maltotrioside, methyl -isopanoside, and methyl -(61--glucopyranosyl)-maltoside, and for the corresponding set of OH-4A, OH-6A, and OH-4B for the nonreducing end -(1,6)-linked methyl -panoside and methyl -isomaltotrioside.

Relatively strong hydrophobic interactions exist with C-6A for all trisaccharides and with C-6B for the nonreducing-end -(1,4)-linked trisaccharides (data not shown). Together with the low flexibility of the -(1,4)-glycosidic bond, no equivalent contribution to the interaction at C-6B exist for other -linked glycosyl disaccharides (Coutinho et al., 1997c), which could to some extent explain the selectivity of GA in digesting disaccharides. As in previous docking studies, strain occurs at the oxygen involved in the glycosidic bond to be cleaved, probably the basis for catalysis (Coutinho et al., 1997b,c). Interestingly, a stronger unfavorable interaction occurs at glycosidic oxygen O-6B involved in branching in methyl -(6¹--glucopyranosyl)-maltoside, suggesting that in the hydrolysis of branched substrates some GA active site residues may have to change position to accommodate the branch and to allow the substrate to bind properly in the catalytic site.

At the third subsite, strong polar interactions are found at OH-6C of productively bound methyl -maltotrioside and methyl -panoside conformers, since it hydrogen-bonds to both the backbone of Gly121 and of Glu179. No significant polar interactions occur for the top optimal structures of the remaining linear trisaccharides.

Tight binding at the first and second subsites occurs only when a glycosyl substrate containing an -(1,4)-linkage is found at the catalytic site. Moreover, the tightest binding is observed for methyl -maltotrioside, the only trisaccharide containing two -(1,4)-linkages. The easier accommodation of this substrate to the active site explains the low values of KM for maltotriose hydrolysis by A. niger GA (Meagher and Reilly, 1989). Steric hindrance at the second subsite is probably the cause of the lower rate observed in the hydrolysis of methyl -(6¹--glucopyranosyl)-maltoside by A. niger GA, when compared to that of methyl b-maltotrioside (Bock, 1987). Similar steric hindrance is found due to the O-6B-linked acetyl group in phenyl 6-O-acetyl--maltoside, that causes a eight-fold increase in KM for a similar kcat when compared to phenyl -maltoside hydrolysis by Rhizopus delemar GA (Hiromi et al., 1973). Even though a productive binding mode has been found for methyl -panoside, the possibility of low-energy unproductive binding modes might explain the same magnitude of KM obtained for panose and isomaltotriose (Meagher and Reilly, 1989). On the other hand, the different clusters found for methyl -isomaltotrioside, even after allowing variation at the third subsite, are all productive at the catalytic site. The extra hydrogen bond found in the third subsite for methyl -panoside might further stabilize the transition state in panose hydrolysis and permit a kcat an order of magnitude higher than for isomaltotriose, which lacks strong interactions there. Furthermore, the presence of an extra -linked glycosyl unit at O-6C in oligosaccharides whose nonreducing end is otherwise identical to that of panose, like 6³,6⁴-di-O-glucopyranosyl maltotetraose and 6³,6⁵,6⁶-tri-O-glucopyranosyl maltohexaose, renders its digestion by GA very difficult (Okada et al., 1994), most likely by making hydrogen bonding at the third subsite impossible.

It is interesting that strong interactions occur between the hydroxymethyl groups of every glycosyl unit in methyl -maltotrioside and the active site, with hydrogen bonding at OH-6A and OH-6C and relatively strong hydrophobic interactions with every C-6. 

An equivalent hydroxyl group involved in binding at the third subsite was described for the binding of maltotetraose-like inhibitors (Aleshin et al., 1994, 1996; Stoffer et al., 1995). This study shows that the nonreducing-end -(1,6)-linked methyl -panoside can bind at the third subsite in a mode surprisingly similar to that of -(1,4)-linked linear substrates. The GA active site, and particularly the second subsite as shown before (Coutinho et al., 1997c), is very adaptable even in its interaction with some longer substrates.

The use of a library of conformers, representing all important conformational states of the substrate, as initial structures for docking assures the exploration of significant conformational space. It demonstrates again the ability of AutoDock to explore and explain important features of substrate flexibility in carbohydrate-protein interactions. Complementary to the crystallographic studies of GA complexes with maltooligosaccharide analogs this study demonstrates both the versatility of the second Aspergillus GA subsite and the role of its third subsite in promoting the hydrolysis of substrates -(1,4)-linked between the second and third subsites. Understanding the molecular basis of Aspergillus GA interaction with different substrates supports current efforts to engineer its selectivity.

The authors gratefully acknowledge the financial support of the U. S. Department of Energy through the Consortium for Plant Biotechnology Research, Inc. and that of Genencor International, Inc. The authors thank Alexander Aleshin and Richard Honzatko for access to crystallographic data prior to public release and to Raymond Lemieux for the suggestions that led to this study.

Aleshin, A. E.; Firsov, L. M.; Honzatko, R. B. Refined Structure for the Complex of Acarbose with Glucoamylase from Aspergillus awamori var. X100 to 2.4-Å Resolution. J. Biol. Chem., 1994, 269, 15631.

Aleshin, A. E.; Stoffer, B.; Firsov, L. M.; Svensson, B.; Honzatko, R. B. Crystallographic Complexes of Glucoamylase with Maltooligosaccharide Analogs: Relationship of Stereochemical Distortions at the Non-reducing end to the Catalytic Mechanism. Biochemistry, 1996, 35, 8319.

Bock, K. Carbohydrate-Protein Interactions: Substrate Specificity of Enzymes used in the Degradation of Oligosaccharides Related to Starch and Cellulose. Pure Appl. Chem., 1987, 11, 1447.

Coutinho, P. M.; Dowd, M. K.; Reilly, P. J. Automated Docking of Monosaccharide Substrates and Analogues and Methyl -Acarviosinide in the Glucoamylase Active Site. Proteins, 1997a, 27, 235.

Coutinho, P. M.; Dowd, M. K.; Reilly, P. J. Automated Docking of Isomaltose Analogues in the Glucoamylase Active Site. Carbohydr. Res., 1997b, 297, 309.

Coutinho, P. M.; Dowd, M. K.; Reilly, P. J. Automated Docking of Glucosyl Disaccharides in the Glucoamylase Active Site. Proteins, 1997c, 28, 162.

Dowd, M. K.; Zeng, J.; French, A. D.; Reilly, P. J. Conformational Analysis of the Anomeric Forms of Kojibiose, Nigerose, and Maltose Using MM3. Carbohydr. Res., 1992, 230, 223.

Dowd, M. K.; Reilly, P. J.; French, A. D. Relaxed-Residue Conformational Mapping of the Three Linkage Bonds of Isomaltose and Gentiobiose with MM3(92). Biopolymers, 1994, 34, 625.

Goodsell, D. S.; Olson, A. J. Automated Docking of Substrates to Proteins by Simulated Annealing. Proteins, 1990, 8, 195.

Goodsell, D. S.; Lauble, H.; Stout, C. D.; Olson, A. J. Automated Docking in Crystallography: Analysis of the Substrates of Aconitase. Proteins, 1993, 17, 1.

Goodsell, D. S.; Morris, G. M.; Olson, A. J. Docking of Flexible Ligands. Applications of AutoDock. J. Mol. Recog., 1996, 9, 1.

Hehre, E. J.; Okada, G.; Genghof, D. S. Configurational Specificity: Unappreciated Key to Understanding Enzymic Reversions and de Novo Glycosidic Bond Synthesis. I. Reversal of Hydrolysis by -, b- and Glucoamylases with Donors of Correct Anomeric Form. Arch. Biochem. Biophys., 1969, 135, 74.

Hiromi, K.; Kawai, M.; Suetsugu, N.; Nitta, Y.; Hosotani, T.; Nagao, A.; Nakajima, T., Ono, S. Kinetic Studies on Glucoamylase. VI. Inhibition of Substrate Analogs. J. Biochem., 1973, 74, 935.

Lemieux, R. U.; Spohr, U.; Bach, M.; Cameron, D. R.; Frandsen, T. B.; Svensson, B.; Palcic, M. M. Chemical Mapping of the Active Site of the Glucoamylase of Aspergillus niger. Can. J. Chem., 1996, 74, 319.

Meagher, M. M.; Reilly, P. J. Kinetics of the Hydrolysis of Di- and Trisaccharides by Aspergillus niger Glucoamylases I and II. Biotechnol. Bioeng., 1989, 34, 689.

Nikolov, Z. L.; Meagher, M. M.; Reilly, P. J. Kinetics, Equilibria, and Modeling of the Formation of Oligosaccharides from D-Glucose by Aspergillus niger Glucoamylases I and II. Biotechnol. Bioeng., 1989, 34, 694.

Okada, Y.; Koizumi, K.; Kitahata, S. Separation and Characterization of Five Positional Isomers of Trimaltosyl-cyclomaltohexaose (Trimaltosyl-b-cyclodextrin). Carbohydr. Res., 1994, 254, 1.

Stoffer, B.; Aleshin, A. E.; Firsov, L. M.; Svensson, B.; Honzatko, R. B. Refined Structure of the Complex of D-gluco-Dihydroacarbose with Glucoamylase from Aspergillus awamori var. X100 to 2.2 Å Resolution: Dual Conformations for Extended Inhibitors Bound to the Active Site of Glucoamylase. FEBS Lett., 1995, 358, 57.