Amylosucrase, a Glucan-synthesizing Enzyme from the α-Amylase Family*

Amylosucrase (E.C. 2.4.1.4) is a member of Family 13 of the glycoside hydrolases (the α-amylases), although its biological function is the synthesis of amylose-like polymers from sucrose. The structure of amylosucrase from Neisseria polysaccharea is divided into five domains: an all helical N-terminal domain that is not similar to any known fold, a (β/α)8-barrel A-domain, B- and B′-domains displaying α/β-structure, and a C-terminal eight-stranded β-sheet domain. In contrast to other Family 13 hydrolases that have the active site in the bottom of a large cleft, the active site of amylosucrase is at the bottom of a pocket at the molecular surface. A substrate binding site resembling the amylase 2 subsite is not found in amylosucrase. The site is blocked by a salt bridge between residues in the second and eight loops of the (β/α)8-barrel. The result is an exo-acting enzyme. Loop 7 in the amylosucrase barrel is prolonged compared with the loop structure found in other hydrolases, and this insertion (forming domain B′) is suggested to be important for the polymer synthase activity of the enzyme. The topology of the B′-domain creates an active site entrance with several ravines in the molecular surface that could be used specifically by the substrates/products (sucrose, glucan polymer, and fructose) that have to get in and out of the active site pocket.

Amylosucrase (AS) 1 is a hexosyltransferase (E.C. 2.4.1.4) produced by non-pathogenic bacteria from the Neisseria genus and was identified in N. perflava as early as 1946 (1). MacKenzie et al. (2) identified intracellular AS in six other Neisseriae species, and later an extracellular N. polysaccharea AS was discovered (3). N. polysaccharea was isolated from the throats of healthy children, and it was suggested that the function of the secreted glucansucrase AS was to produce insoluble polymers. Until recently AS has only been found in bacteria from the Neisseria genus, but the Deinococcus radiodurans genome (4) and the Caulobacter crescentus genome (5) actually encodes proteins with a similar length that are 43 and 34%, respectively, identical to AS from N. polysaccharea.
In the presence of an activator polymer (e.g. glycogen), AS catalyzes the synthesis of an amylose-like polysaccharide composed of only ␣-(134)-glucosidic linkages using sucrose as the only energy source (6). This glycogen pathway is not found in e.g. Escherichia coli, which like most bacteria require activated ␣-D-glucosyl-nucleoside-diphosphate substrates for polysaccharide synthesis (7). The utilization of a readily available substrate makes AS a potentially very useful glucosylation tool for the production of novel amylopolysaccharides.
The recent cloning of N. polysaccharea AS in E. coli (8,9) has made mutational (10) and detailed kinetic studies of highly purified enzyme possible (11,12). It has also provided AS in sufficient amounts for successful crystallization experiments (13). The recombinant AS is derived from a glutathione Stransferase fusion protein and consists of a single polypeptide chain with 636 amino acid residues including 6 cysteines and 15 methionines.
The ␣-amylase reaction mechanism is a general acid catalysis, similar to all of the glucoside hydrolases (16), and the same mechanistic scheme can also accommodate glucan synthesis from sucrose as shown in Scheme 1. The reaction is initiated by simultaneous protonation of the glycosidic bond by a proton donor and a nucleophilic attack on the anomeric carbon of the glucose moiety. This leads to the covalently linked substrateenzyme intermediate. The intermediate can react with either water or with another saccharide molecule, as shown in the scheme. This implicates that the ratio between hydrolysis and transglycosylation is determined only by the relative concentrations of water and sugar moieties in the active site.
Apart from AS, the GH Family 13 comprises other enzymes with non-hydrolytic functions. The crystal structure of a cyclo-dextrin glucanotransferase (17) (CGTase) showed an active site architecture very similar to the ␣-amylase from Aspergillus oryzae (18) (TAKA-amylase, the first ␣-amylase structure determined) and thus implicated very similar reaction mechanisms at least for the formation of the covalent intermediate. The presence of a covalent intermediate has been verified experimentally in CGTases (19).
These findings all suggest that the active site of AS is highly similar to those of the ␣-amylases and CGTases. Structure determinations of complexes with substrate analogues have yielded detailed information on the ␣-amylase structure/function relationships. A stringent nomenclature for enzyme-substrate interactions has been developed, and substrate binding is usually described in terms of numbered sugar binding subsites (20). The catalytic residues are then located between the sugar binding subsites Ϫ1 and ϩ1, when numbering the polysaccharide from the reducing end. In this work, we present the crystal structure of AS at a resolution of 1.4 Å, which represents the first crystal structure of a glucansucrase and is the first structure of a glucan-elongating enzyme from the GH 13 family. The structural alignment of AS and ␣-amylasesubstrate analogue complexes is well suited to provide a basis for the understanding of product profile and substrate specificity observed for AS.

EXPERIMENTAL PROCEDURES
Crystallization-Expression and purification of recombinant AS was performed as described previously (9,12). The production of Se-Met AS and the crystallization conditions (equal amounts of 4 mg/ml protein solution (150 mM NaCl, 50 mM Tris-HCl, pH 7.0, 1 mM EDTA, and 1 mM ␣-dithiothreitol) and reservoir solution (30% polyethylene glycol M r 6000 and 0.1 M HEPES, pH 7.0)) have been published (13).
Data Collection, Structure Determination, and Refinement-All data were collected at the ESRF, Grenoble and were processed and scaled using DENZO and SCALEPACK (21). Multiple wavelength anomalous dispersion (MAD) data were collected at beamline BM 14. Data were collected using energies corresponding to the inflection point and peak of the experimentally determined selenium K edge and a remote high energy wavelength (Table I). All 15 selenium sites were identified when the MAD data were analyzed by the SOLVE program (22). Phases were extended to 1.7 Å using DM (23), and the structure was build with the automated tracing procedure ARP/wARP (24). This tracing located 625 of the total 628 amino acid residues found in the structure. Later a 1.4 Å native data set was obtained at beamline ID 14 EH 1 ( ϭ 0.934 Å), and the structure was refined at this level of resolution. Further rebuilding was done in program O (25), and refinement was performed with the CNS program package (26). A total of 628 amino acid residues, one Tris molecule, one HEPES molecule, a sodium ion, and 751 water molecules were included in the final model. Refinement statistics are listed in Table I. Several patches of elongated electron density that could arise from a polyethylene glycol molecule were not fitted. Thirtyone of the side chains were fitted with two conformations, and for sixteen surface side chains some of the outermost atoms displayed high B-factors. A polymerase chain reaction error was detected in the structure. Surface residue 537 was found to be an Asp/Asn instead of a Gly predicted by the sequence of the native enzyme. The nucleotide sequence of the recombinant DNA identified the residue as an Asp. The stereochemistry of the final model was analyzed by PROCHECK (27): 91% of the residues were found to lie in the most favorable regions of the Ramachandran plot and 8.6% in the additional allowed regions. Only two residues (Glu 344 and Phe 250 ) were found in a generously allowed region. A schematic representation of the enzyme is shown in Fig. 1. The overall B-factor for the protein is 16.4 Å 2 , whereas it is 15.9 Å 2 and 16.5 Å 2 for the main chain and the side chain atoms, respectively. The B-factors of the C␣ atoms are plotted in Fig. 2. The active site is shown in Fig. 3 as an example of the quality of the 1 (2F o -F c ) electron density.
Coordinates-Coordinates have been deposited at the Protein Data Bank (accession code 1G5A).

RESULTS AND DISCUSSION
Description of the Structure-The single polypeptide chain (628 amino acid residues) is folded into a tertiary structure with five domains named N, A, B, BЈ, and C ( Fig. 1). Residues 1-90 comprise the all ␣-helical N-domain. It contains six amphiphilic helices that we have chosen to name nh1 to nh6. The helices consist of the residues Pro 2 -Leu 12 , Thr 16 -Lys 25 , Ser 26 -Pro 41 , Pro 41 -Gly 52 , Leu 57 -Arg 75 , and Ser 77 -Asn 88 as defined by the Kabsch-Sander algorithm (28) in the program PRO-CHECK (27). No known structures or domains were found to be similar to the AS N-domain in a database search with the DALI (29) server. Two helices from the N-domain (nh4 and nh5) are packed against two helices (h3 and h4) from the central (␤/␣) 8barrel (domain A) forming a four-helix bundle. The interface between the helices in the bundle is almost entirely hydrophobic, and no solvent molecules are located in the interface.
Domain A (residues 98 -184, 261-395, and 461-550) is made up of eight alternating ␤-sheets (e1-e8) and ␣-helices (h1-h8) giving the catalytic core (the well characterized (␤/␣) 8 -barrel) common to the GH Family 13 (Figs. 1 and 4). A characteristic of the (␤/␣) 8 -barrel enzymes is that the loop region connecting strands to helices (labeled loop1 to loop8) are much longer on average than those connecting helices to strands. In particular AS has two loops (loop3 and loop7) that are so long that they constitute the separate domains B and BЈ. The positions of the secondary structural elements within the primary structure of domain A are shown in Fig. 4.
Domain B (residues 185-260) contains two short antiparallel ␤-sheets. The inner sheet (relative to the barrel) is formed by two strands (residues 187-190 and 253-256) and the outer sheet is formed by three strands (residues 211-213, 237-240, and 245-248) flanked by two ␣-helices (residues 193-201 and SCHEME 1. General reaction mechanism for hydrolysis and transglycosylation. (Figs. 1 and 4). A B-domain is found in many ␣-amylases. In TAKA-amylase the main structural feature is a short three-stranded antiparallel ␤-sheet. There are also ␣-amylases that do not have a B domain. For example, barley ␣-amylase (30) has a short hairpin at this position.
Domain C is an eight-stranded ␤-sandwich found C-terminal to the (␤/␣) 8 -barrel (residues 555-628). A C domain is found in other ␣-amylases, for example TAKA-amylase and barley ␣-amylase. Several of these domains are found in the CGTases, but so far the functional role of the C domain is unknown. Although the complete AS sequence contains six cysteine residues no disulfide bridges are found in the structure. Some of the cysteines are exposed on the surface, but no tendency to multimerization has been reported.
The C␣ displacement parameters (B-factors) are plotted in Fig. 2. The enzyme displays low overall thermal vibration with a mean B-factor for all atoms of 16.4 Å 2 . The plot shows that especially the region 250 -400 (from the start of h3 to the beginning of the BЈ domain) has low displacement parameters and that the regions of the molecule with the highest displacement parameters are localized far from the substrate binding pocket.
Relation to Family 13 Enzymes-A comparison of the fulllength enzyme with known protein structures using the DALI server (29) showed that the glycoside hydrolase Family 13 exo-acting enzyme oligo-1,6-glucosidase (31) had the highest similarity to AS. A total of 458 C␣ atoms could be superimposed with an rms of 2.7 Å. The superimposable residues are almost all found in the A and C domains. The structural similarity to  and F c, hkl are the observed and calculated structure factor amplitudes. R free is the same as R cryst but calculated over the 5% randomly selected fraction of the reflection data not included in the refinement.
Family 13 ␣-amylases is also high. In particular, the TAKAamylase-acarbose complex (32) had 368 superimposable C␣ atoms with a rms of 2.8 Å. A structural-based sequence alignment between TAKA-amylase, AS, and oligo-1,6-glucosidase starting after the unique N-domain is shown in Fig. 4. The boxed sequence patches represents regions of genuine structural similarity. Because of the high structural similarity the alignment can be used to propose AS-substrate interactions from enzyme-substrate investigations performed on related enzymes.
Active Site Architecture-The general acid residue Glu 328 and the nucleophile Asp 286 have been identified using conventional sequence alignment (9) and mutational studies (10). These results are supported by the structural alignment found in Fig. 4, which shows that the C␣ positions of the two residues coincides with catalytic residues from both TAKA-amylase and oligo-1,6-glucosidase. Asp 286 and Glu 328 are found at the tips of ␤-sheets 4 and 5 in the (␤/␣) 8 -barrel of AS (Fig. 1), as required for Family 13 members. The distance between Asp 286 C␣ and Glu 328 C␣ is 5.4 Å in accordance with AS being an ␣-retaining enzyme (14). A Tris molecule is bound at the active site (Fig. 3) with a short hydrogen bond (2.6 Å) between O␦2 of Asp 286 and one of the Tris oxygens and several hydrogen bonds to surrounding amino acid side chains (Asp 144 , His 187 , Glu 328 , and Arg 509 ). Tris has previously been found to be a very good probe for the active site of ␣-amylases (33).
The TAKA-amylase-acarbose complex (32) has identified a number of enzyme-substrate active site interactions. Around the Ϫ1 subsite the following interactions are reported (conserved residues at an equivalent position in AS given in parentheses): The nucleophile Asp 206 O␦2 (Asp 286 ) is forming a hydrogen bond to the O6 of the I-ring of the modified acarbose (Fig. 5). His 122 (His 187 ) also forms a hydrogen bond to O6I. Arg 204 (Arg 284 ) forms a salt-bridge to the O␦1 of the nucleophile and is very important for the correct positioning of the nucleophile. Arg 204 (Arg 284 ) has an additional weak hydrogen bond to O 2 I. Glu 230 (Glu 328 ) is the general acid/base. It forms a hydrogen bond to the acarbose N. Asp 297 (Asp 393 ) O␦1 and O␦2 forms hydrogen bonds to O 2 I and O3I respectively. Tyr 82 (Tyr 147 ) provides an important stacking platform for the substrate ring at the Ϫ1 position. Finally His 392 (His 296 ) forms a short hydrogen bond to the hydroxyl O of Tyr 82 (Tyr 147 ) an interaction suggested to be pivotal for the positioning of the stacking platform (32). All of these residues can be found at identical C ␣ positions in TAKA-amylase, AS, and oligo-1,6-glucosidase (Figs. 4 and 5). As seen in Fig. 5, the side chains of these residues are found in identical spatial positions as well. Thus ␣-amylases and AS have very similar active site architecture with respect to the immediate surroundings of the scissile bond (subsite Ϫ1). This is in agreement with the general mechanism outlined in Scheme 1. The mechanism for the formation of the covalent intermediate is similar. But how does AS ensure specificity for sucrose as the first substrate, and how does AS  (18); AS, amylosucrase from N. polysaccharea; gluc, oligo-1,6-glucosidase from B. cereus (31). Stretches corresponding to genuine topological equivalence are boxed (38,39). The AS sequence numbers (AS_seq) and the DSSP assignment of secondary structure elements are inserted (taka_ss, AS_ss, and gluc_ss) (28). The eight alternating ␤-sheets (e1-e8) and ␣-helices (h1-h8) of the (␤/␣) 8 -barrel (domain A) are labeled. The two catalytically active residues Asp 286 and Glu 328 are indicated by a ϩ whereas other important residues (His 187 , His 392 , and Asp 393 ) are marked by a #. prevent water from being the second substrate? These questions can be addressed by studying the superposition of AS and TAKA-amylase at the other subsites mapped out by the TAKAamylase-acarbose complex.
Specificity for Sucrose as the First Substrate-TAKA-amylase is an endo-acting enzyme. It has a total of six specific binding sites for linked ␣-(134)-glucosyl moieties. The TAKAamylase residues reported to be involved in enzyme-substrate contacts in subsites ϩ1 and ϩ2 are not structurally conserved in AS. TAKA-amylase His 210 in subsite ϩ1 donating a hydrogen bond from N⑀2 to O5J is a Phe in AS, whereas TAKAamylase Lys 209 with hydrogen bonds to the modified acarbose hydroxyls OK2 and OK3 is an Ala in AS. This suggests that the ϩ1 subsite in AS is modified to accommodate specificity for the furanosyl ring of sucrose.
The TAKA-amylase-2 subsite has been completely disrupted in AS. Asp 144 from loop2 forms a salt bridge with Arg 509 and thus occupies the subsite. An equivalent salt bridge is observed in the exo-acting oligo-1,6-glucosidase. An Ala and an Asp are found in these positions in TAKA-amylase. The salt bridges gives the active site a pocket topology in AS and oligo-1,6glucosidase, in contrast to the cleft observed in TAKA-amylase. The result of the pocket topology is an exo-acting enzyme. The ␣-amylase cleft is closed by residues from domain B, domain BЈ, and loop2. The bottom of the pocket is quite thin-walled with a solvent accessible dent in the protein surface right behind the Asp 144 -Arg 509 salt bridge. Without this blockage the active site topology would be a tunnel. In conclusion, the assumed furanosyl specificity at the ϩ1 site and the salt bridge creates the sucrose specificity in AS.
Oligosaccharides as Second Substrates-The pocket topology in AS greatly reduces the solvent accessibility to the active site. When examining the AS structure superimposed with the TAKA-amylase-acarbose structure it can be seen that the pocket includes the Ϫ1 and ϩ1 subsites. The superposition also suggests that Phe 250 is sandwiching the I-ring of the modified acarbose at the Ϫ1 subsite with Tyr 147 in AS. TAKA-amylase has a Gly at this position, oligo-1,6-glucosidase also have a Phe. This could implicate a more stable covalent intermediate in AS and oligo-1,6-glucosidase compared with ␣-amylases. This in turns could reflect the reduced accessibility of the active site. The intermediate simply has to exist long enough for the fructose to leave the active site and the second substrate (oligosaccharide or water) to enter.
The pyranosyl ring bound at the TAKA-amylase ϩ2 subsite can just be seen in the surface plot (Fig. 6). Compared with the exo-acting hydrolase oligo-1,6-glucosidase the pocket of AS is very narrow leaving little room for water to enter when an oligosaccharide such as elongated acarbose is in the pocket. In the superimposition of acarbose into the AS structure the modified acarbose molecule fills the pocket almost completely. The surface area of the enzyme around the pocket-entrance is however quite open (Fig. 6). In fact several ravines in the surface leads to the pocket. Hence, it could be speculated that the growing glucan polymer is embedded in one ravine, whereas sucrose/fructose approaches/leaves the active site through another ravine. An architecture like this with a number of remote glucose binding subsites (securing a high "effective" concentration of the oligosaccharide chain) could be responsible for the transferase rather than hydrolase activity of AS. The unique AS domain-BЈ is involved in the formation of these ravines (Fig.  6, dark gray surface). Because of the close proximity to the active site and the high content of aromatic residues (7 Phe, 3 Tyr, and 1 Trp out of 54 residues) it is tempting to propose that the BЈ-domain is essential for the binding of the growing glucan polymer. However, this hypothesis has to be tested by experiments involving complex formation with different oligosaccharides.
Calcium Independence-No calcium ions were found in the FIG. 5. Stereo drawing of the superposition of the active sites of AS (black) and the TAKA-amylase-acarbose complex (gray). Atoms in Asp 286 and Glu 328 were used to create the superimposition. The amino acid residue labels are black for AS and gray for TAKA-amylase. The elongated acarbose molecule is for clarity labeled at the ϩ3 and Ϫ3 subsite. Distances (from the superimposition) between the nucleophile (Asp 286 O␦1) and the anomeric carbon at subsite Ϫ1 in acarbose and between the general acid (Glu 328 O⑀2) and acarbose N are shown (in Å). Furthermore, the distance (in Å) in the salt bridge found in AS between Asp 144 O␦2 and Arg 509 N2 is shown.
structure. This also makes AS more similar to oligo-1,6-glucosidase than to TAKA-amylase. However, most of the Ca 2ϩ site arrangement found in many amylases is conserved. In TAKAamylase a calcium ion is hepta-coordinated by oxygen atoms from Asp 175 (O␦1 and O␦2), Asn 121 (O␦1), Glu 162 (backbone O), His 210 (backbone O), and three water molecules. For both AS and oligo-1,6-glucosidase the calcium ion found in ␣-amylases is replaced with a presumable protonated lysine N (Lys 293 in AS and Lys 206 in oligo-1,6-glucosidase). The two side chains involved in calcium binding (Asn 121 and Asp 175 in TAKA-amylase) are structurally conserved (Asn 186 and Asp 256 ) in AS. However, only Asp 256 O␦2 is within hydrogen bonding distance of the TAKA-amylase Ca 2ϩ site. One of the backbone interactions (O from His 210 in TAKA-amylase) is also conserved in AS (Phe 290 ), but the change in side chain disables the interaction between the "calcium" site and the subsite ϩ1 described for TAKA-amylase (32). The last of the three hydrogen bonds found for Lys 293 N comes from a water molecule. The lack of calcium has also been observed in neopullulanase (34) and maltogenic ␣-amylase (35).