|Home | About | Journals | Submit | Contact Us | Français|
Amylosucrases (ASes) catalyze the formation of an α-1,4-glucosidic linkage by transferring a glucosyl unit from sucrose onto an acceptor α-1,4-glucan. To date, several ligand-bound crystal structures of wild-type and mutant ASes from Neisseria polysaccharea and Deinococcus geothermalis have been solved. These structures all display a very similar overall conformation with a deep pocket leading to the site for transglucosylation, subsite −1. This has led to speculation on how sucrose enters the active site during glucan elongation. In contrast to previous studies, the AS structure from D. radiodurans presented here has a completely empty −1 subsite. This structure is strikingly different from other AS structures, as an active-site-lining loop comprising residues Leu214–Asn225 is found in a previously unobserved conformation. In addition, a large loop harbouring the conserved active-site residues Asp133 and Tyr136 is disordered. The result of the changed loop conformations is that the active-site topology is radically changed, leaving subsite −1 exposed and partially dismantled. This structure provides novel insights into the dynamics of ASes and comprises the first structural support for an elongation mechanism that involves considerable conformational changes to modulate accessibility to the sucrose-binding site and thereby allows successive cycles of glucosyl-moiety transfer to a growing glucan chain.
The transfer of glucose onto a growing α-1,4-glucan chain most commonly involves a ‘high-energy’ glucosyl donor in the form of a nucleotide glucosyl derivative. This reaction is found in most organisms (Cid et al., 2002 ). A small group of bacteria are able to perform glucan synthesis through an additional route that utilizes sucrose as a glucosyl donor. This reaction is catalyzed by amylosucrases (ASes; EC 184.108.40.206), which belong to a family consisting mainly of glycoside hydrolases: glycoside hydrolase family 13 (GH13, http://www.cazy.org). Along with other sucrose-specific enzymes, such as the sucrose hydrolases (SHes), ASes cluster into subfamily 4 (Stam et al., 2006 ). Although hydrolysis does occur as a side reaction, the ASes are efficient in glucan synthesis (Potocki de Montalk et al., 2000 ), whereas the SHes are very efficient hydrolases and do not synthesize glucan (Kim et al., 2004 ). Both reactions occur through a covalent glucosyl-enzyme intermediate (Jensen et al., 2004 ), which reacts with water in the case of hydrolysis or an acceptor sugar moiety in the case of transglucosylation. This implicates that differences in active-site topology determine the outcome of the catalyzed reaction. In overall terms, compared with SHes, ASes are able to bind and present the reducing-end hydroxyl group of the glucan chain to the glucosyl intermediate in a way that protects the covalent intermediate from attack by water.
Both ASes and SHes have been well characterized in terms of crystallization and three-dimensional structure (Skov et al., 2000 , 2001 , 2002 ; Mirza et al., 2001 ; Jensen et al., 2004 ; Kim et al., 2008 ; Champion et al., 2009 , 2012 ; Guérin et al., 2012 ). Overall, they consist of five domains (Fig. 1 ): N [residues 1–97 in AS from Neisseria polysaccharea (NPAS)], B (185–260), B′ (395–460) and C (555–628) arranged around a central (β/α)8-barrel termed domain A, which harbours the two catalytic carboxylates involved in the α-retaining mechanism. Furthermore, structures in complex with substrates have provided details of substrate-binding determinants. Carbohydrate moiety-binding subsites are labelled according to the point of cleavage (and in the case of ASes also transfer), which by definition occurs between subsites −1 and +1 (Davies et al., 1997 ). A sucrose complex of an inactive variant of NPAS showed that the substrate was bound at the bottom of a deep pocket (Mirza et al., 2001 ), while a crystal structure of maltoheptaose in complex with an inactive variant of NPAS suggested that the binding of an acceptor glucan chain in subsites +2 and +3 would effectively block the entrance to subsite −1 (Skov et al., 2002 ). This observation has led to the important question of how sucrose enters the binding pocket during elongation. In general, two mechanisms are possible: either the transfer cycle occurs by the release of glucan from AS after elongation or through conformational changes that enable the entry of a new sucrose molecule. All AS structures published to date display the same overall conformation with little significant difference. The apparent lack of structural flexibility is in favour of the first mechanism. However, the enzymes have so far all been crystallized with either a ligand or a ligand-mimicking buffer/crystallization component in subsite −1.
Here, the crystal structure of a ligand-free AS from Deinococcus radiodurans (DRAS) is presented. Compared with the previously determined AS structures, this structure reveals large changes in the active-site topology. These changes leave subsite −1 completely accessible. The large conformational changes can be attributed to an empty subsite −1. These important and novel structural findings shed new light on the glucan-elongation mechanism and support the hypothesis of an elongation mechanism involving enzyme conformational changes to allow the entry of sucrose to the active site.
Escherichia coli BL21 cells containing pGST-DRAS were grown in 2×YT medium supplemented with 100 µg ml−1 ampicillin. Gene expression was induced with IPTG at a final concentration of 5 µM when the OD600 nm reached 0.3. The cells were then grown for 15 h at 291 K and harvested by centrifugation (6500 rev min−1, 15 min, 277 K). The pellets were resuspended to a final OD600 nm of 80 in PBS buffer (140 mM NaCl, 2.7 mM KCl, 10 mM Na2HPO4, 1.8 mM KH2PO4 pH 7.3) and sonicated. The extract was centrifuged at 12 500 rev min−1 for 20 min at 277 K and DRAS was purified to homogeneity using GST-affinity chromatography followed by dialysis and subsequent tag removal using PreScission protease (Amersham) and reverse affinity chromatography as described previously (Potocki De Montalk et al., 1999 ; Pizzut-Serin et al., 2005 ). Notably, DRAS was purified using HEPES buffers, as opposed to NPAS and AS from D. geothermalis (DGAS) where Tris was used in the purification and was subsequently observed to bind as a substrate-mimicking ligand in the crystal structure (Skov et al., 2001 ; Guérin et al., 2012 ).
Purified DRAS was concentrated to 3.3 mg ml−1 in PreScission cleavage buffer (50 mM HEPES pH 7.0, 150 mM NaCl, 1 mM EDTA, 1 mM dithiothreitol). Crystallization took place using the hanging-drop vapour-diffusion method at 277 K by mixing 2.5 µl protein solution with 2.5 µl reservoir solution (1 M sodium acetate, 0.1 M imidazole, 0.1 M LiCl pH 6.5). A solution consisting of 0.7 M sodium acetate, 0.07 M imidazole, 0.1 M LiCl pH 6.5, 30% glycerol was used as a cryoprotectant before single crystals were mounted in cryo-loops and cooled in liquid nitrogen. Data were collected from a single crystal at 120 K on beamline X11 at DESY, Hamburg, Germany. Data processing to 3.15 Å resolution was performed using MOSFLM and SCALA (Leslie, 1992 ; Evans, 1993 ). The crystals belonged to the tetragonal space group P43212 with one DRAS molecule in the asymmetric unit. A homology model of DRAS created using MODELLER (Šali & Blundell, 1993 ) and the structure of DGAS (PDB entry 3ucq; sequence identity 75%; Guérin et al., 2012 ) was used as a search model for molecular replacement using Phaser (McCoy et al., 2007 ). Model building was performed with Coot (Emsley & Cowtan, 2004 ). Refinement was performed using PHENIX (Brunger, 2007 ; Adams et al., 2002 ). Positional refinement with individual B factors in combination with secondary-structure restraints resulted in an R value and an R free of 18.8 and 25.5%, respectively, with bond-length and bond-angle deviations of 0.014 Å and 2.012°, respectively. At this point TLS refinement was included with ten TLS groups (residues 16–62, 63–164, 165–210, 211–267, 268–329, 330–418, 419–460, 461–556, 557–610 and 611–643) and scale factors for X-ray/stereochemistry and X-ray/ADP weights both set to 0.1. Very poor and incoherent density for residues 1–15, 124–151 and unclear density for residues 229–233 and 529–534 was found and these residues were consequently not modelled. 3σ mF o − DF c density for several conserved (in the DGAS structures with PDB codes 3ucq and 3uer; Guérin et al., 2012 ) water molecules was observed; however, none were included in the model as they failed to return 2mF o − DF c density upon inclusion. No residues were found in disallowed regions of the Ramachandran plot (using PROCHECK at the ADIT validation server; http://deposit.rcsb.org/validate/). The final structure exhibited an R value and an R free of 20.0 and 23.6%, respectively, with bond-length and bond-angle deviations of 0.003 Å and 0.743°, respectively. Data-collection and refinement statistics are presented in Table 1 . The coordinates and structure factors have been deposited in the Protein Data Bank as entry 4ays.
Multiple sequence alignments were made using ClustalW2 (Larkin et al., 2007 ), while graphics and structural alignments of DRAS (PDB entry 4ays), DGAS (PDB entry 3ucq; Guérin et al., 2012 ) and NPAS (PDB entries 1g5a and 1mw0; Skov et al., 2001 , 2002 ) were made using PyMOL (v.1.5.0; Schrödinger LLC).
DRAS consists of the five domains usually observed in the ASes: N (residues 16–79 in DRAS), B (175–251), B′ (390–459) and C (563–643) flanking the central domain A (80–173, 252–389 and 460–562) (Figs. 1 a and 1 b). DRAS displays 75 and 43% sequence identity to DGAS and NPAS, respectively. Structural superposition of these two amylosucrases on DRAS gives overall root-mean-square deviations (r.m.s.d.s) of 1.2 Å (509 Cα atoms aligned) and 1.5 Å (472 Cα atoms aligned), respectively. Notably, the DGAS and NPAS structures align with a lower r.m.s.d. (0.9 Å; 462 Cα atoms aligned), reflecting more similar overall conformations, as discussed below. In contrast to NPAS, DRAS and DGAS have recently been shown to form dimers (Guérin et al., 2012 ). However, previous reports on DRAS suggested a monomeric state (Pizzut-Serin et al., 2005 ). The DRAS crystal structure displays a dimeric assembly similar to the DGAS dimer, in both cases with the protomers related by crystallographic symmetry. Despite high sequence conservation between DRAS and DGAS in the seven regions identified as potential dimerization motifs by Guérin et al. (2012 ), the putative DRAS interface is primarily mediated by two regions from domain N, residues Thr22, Leu25, Arg26, Arg29 and Tyr30 and residues Glu73, Leu76 and Leu77, which are involved in hydrophobic interactions with the corresponding residues from the second protomer (Figs. 1 c and 1 d). A single hydrogen bond between Glu73 OE1 and Thr22 OG1 is also observed. Finally, residues Thr585 and Pro584 from domain C are in close proximity to Arg333 and His336 from domain A of the other protomer and may play a minor role in dimer stabilization. Analysis of the putative DRAS dimer interface using the PISA server (http://www.ebi.ac.uk/pdbe/prot_int/pistart.html; Krissinel & Henrick, 2007 ) shows an interface of 2086 Å2 and indicates that the assembly is not very strong. In comparison, the DGAS dimer interface is 2950 Å2. In addition to the abovementioned hydrophobic patch between the N domains that is fully conserved between the two structures, DGAS residues from domains A and C are extensively involved in the formation of the dimer interface. Furthermore, two salt bridges, between Glu25 and Arg74 and between Asp84 and Arg341, which were proposed to be important for the DGAS dimer (Guérin et al., 2012 ) are not observed in DRAS. Although all of these four residues are conserved in DRAS, Glu21 (corresponding to Glu25 in DGAS) is involved in an intramolecular salt bridge with Arg24, while the loop containing Arg333 (corresponding to Arg341 in DGAS) adopts a different conformation to that observed in DGAS. The increased number of polar interactions and the more pronounced shape complementarity observed in the DGAS dimer compared with DRAS are consistent with the higher thermostability of DGAS (Pizzut-Serin et al., 2005 ; Guérin et al., 2012 ).
The structure of DRAS presented here is devoid of any ligand in subsite −1 (Fig. 2 a). This is the first such structure, as other AS structures all contain carbohydrate ligands and/or the substrate-mimicking molecules Tris or glycerol (Skov et al., 2001 , 2002 ; Mirza et al., 2001 ; Jensen et al., 2004 ; Guérin et al., 2012 ; Champion et al., 2012 ). Tris was intentionally avoided in the purification of DRAS in an attempt to obtain a true apo structure. In comparison with previous structures of ASes, which all display a pocket-type topology, the unique ligand-free state of DRAS imposes dramatic differences in the active-site architecture, in particular around subsite −1 (Fig. 3 ). A prototypical GH13 −1 subsite is optimized to facilitate the formation of the covalent glucosyl-enzyme intermediate, and in the case of amylosucrases also to efficiently exclude water in order to facilitate transferase activity. A general feature of subsite −1 is a Tyr residue that provides a stacking platform for the covalent intermediate (Jensen et al., 2004 ). In the DGAS structure (Guérin et al., 2012 ) this Tyr residue (Tyr140) is located in a loop region between strand 2 and helix 2 of the A domain (Fig. 3 a). Another highly important residue in this region is the conserved Asp137, which hydrogen bonds to OH4 of the glucosyl moiety and at the same time engages in a salt bridge with the conserved Arg520. This salt bridge is important for the ‘exo’-type activity of ASes, as subsites −2, −3 etc. cannot be formed. In all AS structures elucidated so far, subsite −1 has been situated in a deep pocket. Compared with GH13 hydrolases, for example α-amylase, subsite −1 in ASes is effectively closed by an extended loop of the B domain (Fig. 3 a). As can be seen from the DGAS structure, two Phe residues from this loop are introduced into the active-site pocket: Phe222 is involved in stabilization of the glucosyl moiety through van der Waals interactions, while Phe225 blocks access of solvent from the nonreducing end. Both Phe residues are conserved in DRAS. In NPAS Phe225 is replaced by a glutamine residue (Gln232) that similarly seals subsite −1 through a hydrogen bond to Asp506.
Most surprisingly, it was found that very weak and incoherent density could be observed for residues 124–151 of DRAS corresponding to the loop region between strand 2 and helix 2. Excluding bulk-solvent correction and anisotropic scaling during map calculation returned more 2mF o − DF c density for a number of the missing residues (Figs. 1 a and 1 b); however, it remained impossible to model this region. These residues were excluded from the final model. While proteolytic trimming of the protein cannot be excluded, the presence of electron density in this region, although weak, strongly suggests rather that the protein is intact but highly flexible. To our knowledge, all other GH13 structures determined to date are structurally well defined in this region. Since Asp133 and Tyr136 (corresponding to Asp137 and Tyr140 in DGAS) are part of this disordered region, subsite −1 is only partially formed in the DRAS structure and there is no salt bridge to terminate the binding site after subsite −1 (Fig. 3 b). The remaining residues of subsite −1 are well defined and exhibit good 2mF o − DF c density (Fig. 2 a). Moreover, the B′ domain of DRAS is shifted approximately 4 Å away from the active site and the extended loop of the B domain (residues Leu214–Asn225) adopts a previously unseen conformation (Figs. 2 b and 3 b), leaving subsite −1 completely solvent-exposed. In this conformation, the loop interacts with residues Pro435 and Gln436 from the same protomer and with several residues from the N domain of a symmetry-related molecule. Together, the unique structural features observed in the ligand-free DRAS result in a slightly wider active site displaying a groove-like topology and a fully accessible sucrose-binding site (Fig. 4 ).
SH from Xanthomonas campestris is the only other structure of a GH13 subfamily 4 enzyme which has been determined with an empty −1 subsite (Champion et al., 2009 ). In this structure, the extended loop from the B domain was completely disordered and part of the region corresponding to DRAS residues 124–154 had moved compared with the complexed structures of NPAS (Jensen et al., 2004 ; Skov et al., 2001 , 2002 ; Mirza et al., 2001 ), allowing access to subsite −1 from the nonreducing end. Based on this structure, it was speculated that the extent of movement observed was to be expected, since this enzyme is primarily a hydrolase. DRAS, on the other hand, has clearly been shown to be an AS (Pizzut-Serin et al., 2005 ). The structure presented here therefore suggests that ASes also have highly flexible regions in the absence of bound substrate. This finding is further supported by recent results from MD simulations of DGAS and NPAS that indicated significant mobility of several loops surrounding the active site, in particular the extended loop of domain B, as well as disruption of the salt bridge in subsite −1 in unliganded NPAS (Guérin et al., 2012 ; Champion et al., 2012 ).
The DRAS structure reveals a high degree of flexibility in the substrate-binding region. This is probably also the case for other ASes, considering the relatively conserved nature of the regions involved (Guérin et al., 2012 ). While the sucrose-binding subsites −1 and +1 are only partially formed in the ligand-free structure, the acceptor glucan subsites +2, +3 and +4 appear to be largely unchanged; i.e. a bound glucan at these subsites would be able to make most of the same interactions with the enzyme in both the open (DRAS) and the closed (NPAS and DGAS) forms. Therefore, it is likely that the glucan is able to remain attached at these sites in both conformations. Additionally, AS is expected to be in close proximity to glucan in vivo, as two additional glucan-binding sites distant from the active site have previously been observed on the surface of NPAS (Skov et al., 2002 ). Moreover, the presence of glycogen has an activating effect on NPAS (Potocki De Montalk et al., 2000 ). However, based on previous AS structures, which all display the closed conformation, the presence of a glucan at the reducing-end subsites appeared to efficiently block entry to and exit from the sucrose-binding site (Skov et al., 2002 ), implying the necessity of glucan release between successive glucosyl-transfer reactions. The major structural changes observed in the ligand-free DRAS structure, however, suggest a much more dynamic nature of the ASes, with considerable conformational changes in the loop regions regulating access to the sucrose-binding site. Such a mechanism would allow multiple cycles of glucosyl transfer onto a growing polymer chain without the need for glucan dissociation from the AS during elongation. Furthermore, it provides a plausible pathway for the escape of the fructose moiety that is released during the transglucosylation reaction, another issue that has so far remained unexplained.
Kristoffer Rosenstand is greatly acknowledged for excellent technical assistance and the EU project CEGLYC for financial support. EMBL Hamburg is thanked for beam time.