New evidence for the role of calcium in the glycosidase reaction of GH43 arabinanases Daniele de Sanctis 1,2, *, Jose ´ M. Ina ´ cio 1, * , , Peter F. Lindley, Isabel de Sa ´ -Nogueira 1,3 and Isabel Bento 1 1 Instituto de Tecnologia Quı ´ mica e Biolo ´ gica, Universidade Nova de Lisboa, Oeiras, Portugal 2 Structural Biology Group, European Synchrotron Radiation Facility, Grenoble, France 3 Departamento de Cie ˆ ncias da Vida, Faculdade de Cie ˆ ncias e Tecnologia, Universidade Nova de Lisboa, Caparica, Portugal Keywords Bacillus subtilis; catalytic mechanism; crystallography; endo-a -L-arabinananase GH43; mutagenesis Correspondence I. Bento, Instituto de Tecnologia Quı ´ mica e Biolo ´ gica, Universidade Nova de Lisboa, Avenida de Repu ´ blica-EAN, 2780-157 Oeiras, Portugal Fax: +351 21 441 1277 Tel: +351 21 446 9100 E-mail: bento@itqb.unl.pt I. de Sa ´ -Nogueira, Instituto de Tecnologia Quı ´ mica e Biolo ´ gica, Universidade Nova de Lisboa, Avenida de Repu ´ blica-EAN, 2780-157 Oeiras, Portugal Fax: +351 21 441 1277 Tel: +351 21 446 9100 E-mail: sanoguei@itqb.unl.pt Present address Instituto de Biotecnologia e Bioengenharia- Centro de Biomedicina Molecular e Estrutural, Universidade do Algarve, Campus de Gambelas, Faro, Portugal *These authors contributed equally to this work Database Structural data for the native BsArb43B, the BsArb43B H318A mutant and the BsArb43B D171A mutant in complex with arabinohexose have been submitted to the Protein Data Bank under the accession num- bers 2X8F, 2X8T and 2X8S, respectively (Received 13 May 2010, revised 27 July 2010, accepted 6 September 2010) doi:10.1111/j.1742-4658.2010.07870.x Endo-1,5-a-l-arabinanases are glycosyl hydrolases that are able to cleave the glycosidic bonds of a-1,5-l-arabinan, releasing arabino-oligosaccharides and l-arabinose. Two extracellular endo-1,5-a-l-arabinanases have been isolated from Bacillus subtilis, BsArb43A and BsArb43B (formally named AbnA and Abn2, respectively). BsArb43B shows low sequence identity with previously characterized 1,5-a-l-arabinanases and is a much larger enzyme. Here we describe the 3D structure of native BsArb43B, biochemical and structure characterization of two BsArb43B mutant proteins (H318A and D171A), and the 3D structure of the BsArb43B D171A mutant enzyme in complex with arabinohexose. The 3D structure of BsArb43B is different from that of other structurally characterized endo-1,5-a-l-arabinanases, as it comprises two domains, an N-terminal catalytic domain, with a 3D fold similar to that observed for other endo-1,5-a-l-arabinanases, and an additional C-terminal domain. Moreover, this work also provides experi- mental evidence for the presence of a cluster containing a calcium ion in the catalytic domain, and the importance of this calcium ion in the enzy- matic mechanism of BsArb43B. Abbreviations ABN, arabinanase; AFN, arabinofuranosidase; APBS, adaptive Poisson-Boltzmann solver; CBM, carbohydrate- binding module; GH, glycoside hydrolase; MPD, 2-methyl-2,4-pentadiol; SAD, single wavelength anomalous dispersion; Se-Met, Se- Methionine; TLS, translation/libration/ screw. 4562 FEBS Journal 277 (2010) 4562–4574 ª 2010 The Authors Journal compilation ª 2010 FEBS Introduction The plant cell wall is structurally complex and bio- logically recalcitrant. Micro-organisms, in particular saprotrophs, play a fundamental role in the decompo- sition processes of plant biomass, secreting numerous polysaccharide-degrading enzymes that attack cellu- lose, hemicellulose and pectin. Mobilization of plant biomass for chemical and fuel production is a major biotechnological challenge of the 21st century, and the use of polysaccharide-hydrolysing enzymes in biomass saccharification is promising [1,2]. Although recent years have seen significant advances in interpretation of new structures of (hemi)cellulose hydrolytic enzymes, a full understanding of the details of sub- strate recognition and catalysis by these varied and highly specific enzymes remains an important goal [3]. Hemicellulose is the second most abundant renew- able biomass polymer after cellulose. This fraction of plant cell walls comprises a complex mixture of poly- saccharides that includes xylans, arabinans, galactans, mannans and glucans. l-arabinose, the second most abundant pentose in nature, is found in significant amounts in homopolysaccharides, branched and de-branched arabinans, and heteropolysaccharides such as arabinoxylans and arabinogalactans. Arabinan is composed of a-1,5-linked l-arabinofuranosyl units, some of which are substituted with a-1,3- and a-1,2- linked chains of l-arabinofuranosyl residues [4,5]. Two major enzymes hydrolyse arabinan: a-l-arabinofurano- sidases (AFNs; EC 3.2.1.55) and endo-1,5-a-l-arab- inanases (ABNs; EC 3.2.1.99). AFNs catalyze the hydrolysis of terminal non-reducing a-l-1,2-, a-l-1,3- and a-l-1,5-arabinosyl residues from various oligosac- charides and polysaccharides, including arabinan, ara- binoxylan and arabinogalactan [6,7]. ABNs attack the glycosidic bonds of the a-1,5-l-arabinan backbone, releasing a mixture of arabinooligosaccharides and l- arabinose [4]. These types of enzyme have attracted much attention due to their application in various fields such as food technology, nutritional medical research, plant biochemistry and organic synthesis [4,5,8]. Bacillus subtilis, a saprophytic Gram-positive endospore-forming bacterium, which is a commonly used micro-organism in the antibiotic and enzyme pro- duction industries, synthesizes two AFNs, encoded by the genes abfA and abf2, and two endo-ABNs, BsArb43A and BsArb43B, which are the products of abnA and abn2 genes, respectively. Recently, the four Bacillus subtilis arabinases were independently charac- terized at the genetic and biochemical level [9–12]. Both the endo-ABNs, BsArb43A and BsArb43B, belong to glycoside hydrolase (GH) family 43, a heter- ogeneous group of enzymes comprising endo- and exo- a-l-arabinanases (EC 3.2.1.99), b-xylosidases (EC 3.2.1.37), a-l-arabinofuranosidases (EC 3.2.1.55), xylanases (EC 3.2.1.8) and galactan 1,3-b-galactosidas- es (EC 3.2.1.145) (http://www.cazy.org/) [13]. The crys- tal structures of an exo -ABN, CjArb43A from Cellvibrio japonicus [14] and three endo-ABNs, BsArb43A from B. subtilis [15], ABN-TS from Bacil- lus thermodenitrificans TS-30 [16] and AbnB from Geo- bacillus stearothermophilus [17], have been determined, and showed a catalytic domain consisting of a five- bladed b-propeller fold. However, BsArb43B is a much larger enzyme, and displays less than 23% amino acid identity with previously characterized ABNs. We have previously reported the crystallization and preliminary X-ray analysis of BsArb42B [18]. Here we present the 3D structure of the wild-type enzyme and describe mutant proteins, providing new evidence for the roles of the calcium cluster observed in the active cleft and particular amino acids in enzymatic activity. Results and Discussion BsArb43B (Abn2) structure The three dimensional structure of BsArb43B comprises all 443 amino acids of the mature protein. BsArb43B consists of two domains, an N-terminal catalytic domain (Ala28–Tyr367) and a C-terminal domain (Ala368–Ala470). The catalytic domain displays a char- acteristic b-propeller fold [19,20], with five b-sheets, called blades, arranged radially around a pseudo five- fold axis (Fig. 1). Each blade comprises four anti- parallel b-strands, and the catalytic domain comprises 20 b-strands and three a-helices. Two a-helices are located after blade I, while the third is observed in a coil region between the third and the fourth b-strands of blade IV (Fig. 1). In BsArb43B, a connection between the N- and C-terminal domains is made from the last blade through a long linker, making the last b-strand of this blade much shorter than the other strands. The extra C-terminal domain comprises eight anti-parallel b-strands and a small a-helix, arranged in a b-barrel-like fold (Fig. 1). The catalytic domain The BsArb43B catalytic domain has the b-propeller fold that is characteristic of this type of enzymes. Superposition of the Ca trace of the BsArb43B D. de Sanctis et al. Role of calcium in the glycosidase reaction FEBS Journal 277 (2010) 4562–4574 ª 2010 The Authors Journal compilation ª 2010 FEBS 4563 catalytic domain with the Ca trace of BsArb43A from B. subtilis (Protein Data Bank code 1UV4 [15]), a-l- arabinanase (CjArb43A) from C. japonicus (Protein Data Bank code 1GYD [14]), endo-1,5-a-l-arabinanase (ABN-TS) from B. thermodenitrificans TS-30 (Protein Data Bank code 1WL7 [16]) and endo-1,5-a-l-arab- inanase (AbnB) from Geobacillus stearothermolhilus (Protein Data Bank code 3CU9 [17]) using the LSQKAB program [22] gave the following rmsd val- ues: 1.50 A ˚ for 192 Ca pairs, 1.66 A ˚ for 189 Ca pairs, 1.68 A ˚ for 211 Ca pairs and 1.66 A ˚ for 189 Ca pairs, respectively (Fig. 2). These values show that the sec- ondary structure of the b-propeller is well conserved, with the differences located mainly in the coil regions that connect the five blades, and in the region that connects the N- and C-terminal domains within the fifth blade (Fig. 2). As described above, BsArb43B has two a-helices after blade I, where usually only one is observed, and a third a-helix in blade IV that is not observed in the other endo-arabinanases. The BsArb43B catalytic domain does not show the ‘Velcro’ closure [19,23,24] that is characteristically observed in other proteins with the b-propeller fold. In the ‘Velcro’ closure, the N- and C- termini are joined in the same sheet to ‘seal’ the circular array of the b-propeller [20]. In BsArb43B, closure of the b-propel- ler is achieved in a different way, by a set of polar and hydrophobic interactions established within the N-ter- minal catalytic domain and between the N- and C-ter- minal domains (Fig. 1). These types of interactions are observed either between residues located in blades IV and V and the C-terminal domain or between the hair- pin that joins blades IV and V and the C-terminal domain (Fig. 1 and Table S1). The apolar interactions in the interface between the two domains include the following residues: His37, His355, His345, Val36, Pro39, Ile41, Phe48, Val50, Leu63, Trp66, Tyr322, Tyr331, Ile333, Val347 and Val 349, while the polar interactions are mainly hydrogen bonds and are listed in Table S1. In a similar manner to the other members of the GH43 family, the BsArb43B catalytic domain contains a large cavity that extends across the protein. During refinement of the structural model, additional electron density was observed in this cavity close to the catalytic site, which could not be accounted for by protein atoms. This density was modelled as a metal ion, here refined as a calcium ion, hepta-coordinated by six water molecules and a histidine residue (His318), giving a cluster with a pentagonal bi-pyramid shape (Fig. 3A). BsArb43B active site The active site of BsArb43B is located in the deep cav- ity at the centre of the b-propeller and comprises three Fig. 1. Three-dimensional structure of BsArb43B created using PyMol [21]. The N-terminal catalytic domain comprises a five-blade b-propeller [blade I (residues 40–44, 47–51, 57–60, 67–70) shown in orange; blade II (residues 103–106, 112–119, 126–133, 141–149) shown in magenta; blade III (residues 173–176, 182–186, 193–197, 213–216) shown in blue; blade IV (residues 223–231, 236–243, 253–259, 295–298) shown in dark red; blade V (residues 312–323, 330–337, 346–354, 360–362) shown in cyan] and three short a-heli- ces. Regions connecting the blades are shown in green. The C-ter- minal domain comprises eight b-strands arranged in a distorted b-barrel-like configuration (shown in yellow). The hairpin that joins blades IV and V in the N-terminal domain and interacts with the C-terminal domain is shown in red. Fig. 2. Structural overlay with other arabinanases (EC 3.2.1.99) belonging to the GH43 family. BsArb43B is shown in blue, Cellvib- rio japanicus exo-arabinanase (Protein Data Bank code 1GYD) in red, Bacillus subtilis arabinanase (Protein Data Bank code 1UV4) in green, Bacillus thermodenitrificans arabinanase (Protein Data Bank code 1WL7) in orange, and arabinanase from Geobacillus ste- arothermolhilus (Protein Data Bank code 3CU9) in yellow. Role of calcium in the glycosidase reaction D. de Sanctis et al. 4564 FEBS Journal 277 (2010) 4562–4574 ª 2010 The Authors Journal compilation ª 2010 FEBS carboxylate residues: Asp38, Asp171 and Glu224 (Fig. 3A). These three acidic residues, conserved in all members of GH families 32, 43, 62 and 68 [25], are responsible for the general acid catalysis that leads to hydrolysis of the glycosidic bond. Within GH family 43, the enzymes work by an inversion mechanism, in which one carboxylate acts as a general base catalyst, deprotonating the nucleophilic water molecule that attacks the bond, and the second acidic residue acts as a general acid catalyst, protonating the departing agly- cone [26]. Putative roles of the third residue include acting as a pK a modulator and maintaining the correct alignment of the general acid residue relative to the substrate [14,27]. In the case of BsArb43B, the general base and the general acid (OD1 Asp38 and OE2 Glu224, respectively) are located approximately 5.8 A ˚ apart, and the third catalytic carboxylic acid (Asp171) is located 4.1 A ˚ from the general acid. To probe the function of the three residues, each of them was inde- pendently substituted by an alanine. The arabinanase activity of the mutants D38A, D171A and E224A was assayed in an Escherichia coli periplasmic fraction and compared to that of the wild-type (WT). Under these conditions, the mutants displayed no measurable activ- ity (data not shown), confirming the key roles of each member of the triad of carboxylates in the catalytic activity of BsArb43B. As described above, a metal ion was observed fur- ther down in the catalytic cavity, approximately 5 A ˚ below the catalytic carboxylates, which was hepta- coordinated to six water molecules and a histidine ligand. The presence of ions in an equivalent location has been previously reported for other arabinanases structures. In the ABN-TS model, a chloride ion was assigned to this site [16]. A chloride was also indicated for CjArb43A, but the authors did not exclude the possibility that a calcium ion was present [14], and two Ca 2+ ions were modelled with in the axial cavity of BsArb4A [15]. Recently, a calcium ion was also modelled at this position in the structure of the endo- arabinanase from G. stearothermophilus [17]. In the BsArb43B structure, the coordination distances and geometry at this site strongly suggest the presence of a Ca 2+ ion in the axial cavity (Table 1). This was first confirmed by an X-ray fluorescence spectrum on a native crystal, for which a peak corresponding to cal- cium was observed (Fig. S1), even when no calcium salt was added to the protein solution during purifica- tion or crystallization. In addition, a 12r peak was observed at that position on an anomalous difference Fourier map, calculated from data collected at 1.067 A ˚ wavelength, and was refined perfectly as a calcium ion (Fig. 3B). However, a chloride peak was also observed in the X-ray fluorescence spectrum, and chloride is present in the protein buffer (NaCl + Tris ⁄ HCl) and the crystallization solution [18]. However, chloride ions do not normally adopt such a coordination, but it is typical of calcium ions (Table 1) [28], and would not result in such a high peak of density in the anomalous A B Fig. 3. (A) Detail of the BsArb43B active site showing the three catalytic carboxylates (D38, D171 and E224), the Ca 2+ cluster and the Tris molecule observed in the binding pocket. The water mole- cules are represented by red spheres. The anomalous Fourier map that corresponds to the Ca 2+ ion is shown by a green mesh, and is contoured at the 5r level. (B) Detail of the active site of the BsArb43B H318A mutant (shown in dark grey) superposed on native Abn2 (coloured according to the atom type). The cluster undergoes a small reorganization, and a more planar conformation of the five water molecules and the metal ion is observed. D. de Sanctis et al. Role of calcium in the glycosidase reaction FEBS Journal 277 (2010) 4562–4574 ª 2010 The Authors Journal compilation ª 2010 FEBS 4565 Fourier map at this wavelength. Together, these results confirm that the atom in the cluster is calcium. To determine whether the calcium ion has a specific role in the activity of BsArb43B, assays were per- formed in the presence of EGTA, a chelating agent that binds Ca 2+ with a significantly larger affinity than EDTA does. The results revealed a drastic decrease in the activity of the enzyme in the presence of 1 mm EGTA (14.94 ± 2.93 UÆmg )1 ), compared with the activity values in the presence of EDTA (86.18 ± 13.78 UÆmg )1 ) or in the absence of chelators (90.94 ± 7.85 UÆmg )1 ). These results suggest that the presence of calcium is important for the optimal activity of BsArb43B. Furthermore, to determine whether the role of the calcium is structural, thermal shift assays were per- formed to determine the T m of the protein in the pres- ence and absence of the calcium ion, using EDTA and EGTA as chelators, as described above. The sam- ples were incubated with both chelators for 24 h prior to the thermal shift assays being performed. The T m values obtained for the native enzyme and for the samples incubated with EGTA and EDTA differ by 2 °C and 1 °C, respectively. In addition, whereas the native and EDTA samples show the same type of sharp transition, the transition in the sample incu- bated with EGTA is much smaller, with a broader minimum (Fig. S3). These results indicate that the protein is less stable when incubated with EGTA. As EGTA is a strong calcium chelator, it can be postu- lated that the loss of stability is associated with loss of the calcium cluster. As described above, the calcium atom is coordinated by six water molecules and a histidine residue (His318) (Fig. 3A). Moreover, the water molecules all lie within hydrogen bonding distance of oxygen carbonyl atoms from the protein main chain (Fig. 3B). It is therefore not surprising that the calcium cluster contributes to overall stabi- lization of the b-propeller fold. To further investigate the importance of the calcium cluster, two mutants were produced in which the histi- dine residue that coordinates the calcium was mutated into an alanine (H318A) and a glutamine (H318Q). These mutations aimed to disrupt the calcium cluster in order to determine its importance for this type of pro- tein. In enzymatic assays performed with both mutants, there was a drastic decrease in enzymatic activity is observed for the H318A mutant, and enzymatic activity was completely lost for the H318Q mutant (Table 2). Structure determination of the H318A mutant showed essentially the same structure as that for the native enzyme, with a rmsd of 0.20 A ˚ for 442 Ca pairs. How- ever, in this mutant, a major difference was observed in coordination of the calcium cluster. Surprisingly, muta- tion of His318 to Ala does not unduly disrupt the cluster, and hepta-coordination of the calcium ion is maintained by an extra water molecule that is posi- tioned where the NE2 of the histidine imidazole ring would be located (Fig. 4C). Removal of the histidine residue has the effect of relaxing the geometry of the cluster, resulting in a more planar arrangement of the Ca 2+ ion with five of the water molecules. As stated above, both mutations affect the enzy- matic activity of the mutant proteins, with a complete Table 1. Coodination distances of the calcium cluster. BsArb43B D171A BsArb43B H318A Native BsArb43B Distance to Ca atom (A ˚ )B factor (A ˚ 2 ) Distance to Ca atom (A ˚ )B factor (A ˚ 2 ) Distance to Ca atom (A ˚ )B factor (A ˚ 2 ) Ca 5.52 9.28 11.84 NE2 His318 2.54 7.26 2.53 9.55 w 1 2.52 7.28 2.49 7.53 2.45 11.17 w 2 2.47 8.27 2.60 11.51 2.44 9.37 w 3 2.48 7.22 2.48 11.11 2.50 9.58 w 4 2.46 6.05 2.54 8.10 2.52 10.58 w 5 2.44 5.61 2.57 6.28 2.50 10.72 w 6 2.43 6.97 2.49 6.49 2.42 10.03 w 7 2.49 10.96 Table 2. Catalytic activity of wild-type and mutants of BsArb43B. Kinetic parameters were determined using linear arabinan as the substrate. NA, no detectable activity. Enzyme k cat (s )1 ) K m (lM) k cat ⁄ K m (s )1 lM )1 ) BsArb43B (WT) 191.6 ± 5.9 111.0 ± 3.0 1.72 BsArb43B H318A 3.1 ± 0.5 76.0 ± 3.6 0.04 BsArb43B H318Q NA NA NA Role of calcium in the glycosidase reaction D. de Sanctis et al. 4566 FEBS Journal 277 (2010) 4562–4574 ª 2010 The Authors Journal compilation ª 2010 FEBS loss of activity in the H318Q mutant. As the calcium cluster is present in the H318A mutant, it appears that the decrease in activity observed in this mutant may be due to absence of the histidine residue. This histidine residue is conserved in the majority of arabinanases (Fig. S2), and a histidine residue was also found in an equivalent position in the structure of a b-xylosidase from the GH43 family. For this b-xylosidase from G. stearothermophilus, it was suggested that the histi- dine residue was involved in substrate recognition by establishing a hydrogen bond with a unit of xylose in a substrate–enzyme complex [27]. Likewise, in the structure of native BsArb43B, the histidine residue (His318) also has the ND1 atom within hydrogen bonding distance of a Tris molecule which has been modelled in the active site. Superposition of the struc- tural models for BsArb43B and 2EXJ shows that the Tris molecule is located where the xylose molecule is observed in the b-xylosidase. These observations sug- gest that the histidine residue is also involved in sub- strate recognition and stabilization in BsArb43B. In the absence of the histidine residue, recognition and stabilization of the substrate are not as efficient as for the native enzyme, and the efficiency of the enzyme therefore decreases. On the other hand, when the histi- dine residue is mutated into a glutamine, not only are recognition and stabilization of the substrate compro- mised, but there may also be disruption of the calcium cluster, with a concomitant complete loss of activity, as observed. It is probable that the calcium ion con- tributes to modulation of the pK a values of the cata- lytic carboxylates, thus ensuring the protonation equilibrium necessary for enzyme activity. Substrate-binding cleft Analysis of the molecular surface of BsArb43B shows an elongated surface groove across the face of the pro- peller, which acts as the substrate-binding cleft (Figs 4 and 5). This type of cleft, open on both sides, can accommodate several sugar units of a polymeric sub- strate, and has been observed in other structurally characterized endo-arabinanases [14–17]. The sugar- binding cleft traverses the catalytic domain, and is located between blades II and III on one side and blades V and I on the other. Structural comparison of the BsArb43B binding cleft with the binding cleft of the other four arabinanases with known structure (BsArb43A, ABN-TS, AbnB and CjArb43A; see above for details) shows that differences are mainly found in three loop regions (Fig. 5). Loop region I is located in one side of the binding cleft, comprises residues 53–55, and aligns with the loop that includes residues 30–35 A B B Fig. 4. (A) Surface charge distribution of BsArb43B in the proximity of the active site. Arabinotriose in the binding cleft is represented using sticks. (B) Detailed view of the binding cleft with arabinotri- ose. (C) Details of the residues defining the binding cleft. Hydrogen bond interactions between arabinotriose and the polar residues of the cleft are shown as dashed lines. D. de Sanctis et al. Role of calcium in the glycosidase reaction FEBS Journal 277 (2010) 4562–4574 ª 2010 The Authors Journal compilation ª 2010 FEBS 4567 (amino acid sequence LTEER) in BsArb43A. In this latter enzyme, the loop was suggested to induce endo- activity in the exo-arabinanase CjArb43A from C. japonicus [15]. However, in BsArb43B and the other two endo-arabinanases, this loop is more similar in size to that found in the exo-ABN, and does not show the same sequence motif. These observations indicate that this loop is not in itself responsible for this type of enzymatic activity. The loop II region comprises resi- dues 227–233 in BsArb43B, and is similar in the four arabinanases BsArb43B, ABN-TS, CjArb43A and AbnB, but not BsArb43A. BsArb43A has a longer loop located in a different position (Fig. 5). Loop III is located in the other side of the binding cleft and comprises residues 279–286 in BsArb43B (Fig. 5). This loop is of similar size in all four endo-arabinanases (BsArb43B, ABN-TS, BsArb43A and AbnB), but is much longer in the exo-ABN (CjArb43A) and blocks one of the ends of the binding cleft (Fig. 5). In fact, in the structure of the exo-ABN complex with arabinohexose [15], this loop makes the reducing end of the carbohydrate chain bend towards the solvent, probably optimizing binding of the carbohydrate chain in the cleft, and resulting in trioses as products. These observations suggest that, by blocking one of the ends of the binding cleft, loop III in exo-ABN may be asso- ciated with the exo activity observed in this enzyme, whereas a much shorter loop, as observed in the endo-arabinanases BsArb43B, BsArb43A, AbnB and ABN-TS), which leaves this side of the binding cleft open, is more suitable to accommodate a long poly- meric carbohydrate chain [16]. In order to identify the residues involved in sub- strate recognition, the structure of the D171A BsArb43B mutant in complex with arabinohexaose was determined. Some electron density was found in the proximity of the catalytic site that could be mod- elled as an arabino-trisaccharide (Fig. 4). The surface of the cleft is defined by residues Trp100, Cys119, Pro124, Tyr189, His220, Leu246, Phe284 and Phe285, and by the main chain of residues Asp122 and Ser123. The trisaccharide molecule was found to interact directly with the enzyme by establishing hydrogen bonds with residues Ser190, His220 and Glu224, and indirectly, through water molecules, with residues Asn166, Ser188, Tyr189 and Leu246. These interac- tions are established by the two more deeply buried arabinose residues (AHR2 and AHR3). Stacking inter- actions were also observed between these two arabi- nose rings and Tyr189 and Pro124, respectively. No interaction was observed between the most exposed arabinose residue of the trisaccharide (AHR1) and the protein surface. This arabinose residue is oriented towards the solvent region, and presumably the other residues of the arabinohexose are disordered and do not interact directly with the protein (Fig. 4B). The location of the trissacharide residue in the BsArb43B D171A mutant is similar to that observed for AbnB of G. stearothermophilus in complex with arabinotriose (Protein Data Bank code 3D5Z [17]). However, in the latter structure, the arabinotriose is more buried in the binding pocket, occupying subsites )1 to +2, and the first arabinose residue interacts directly with the catalytic residues Asp27 and Asp147, adopting an equivalent position to that observed for a Tris molecule in the native BsArb43B structure. Indeed, in BsArb43B D171A complex structure, the most internal arabinose ring (AHR3) is located approximately 3.5 A ˚ closer to one of the catalytic resi- dues (Glu224), positioning the arabinose rings AHR3 and AHR2 in positions equivalent to subsites +1 and +2 of the arabinotriose saccharide observed in the G. stearothermophilus AbnB complex. The fact that a fully occupied arabinose residue was not observed in a position equivalent to the )1 subsite in the D171A mutant structure was interpreted as a consequence of a strongly reduced affinity of the mutant enzyme for binding arabinose due to mutation of the catalytic resi- due Asp171 into an alanine. This is further supported by the fact that, even in the structure of the D171A Fig. 5. Molecular surface of BsArb43B showing the loops that dif- fer between the endo- and exo-arabinanases highlighted in different colours: blue for BsArb43B, magenta for BsArb43A, green for ABN- TS, dark grey for AbnB, red for exo-Abn. Role of calcium in the glycosidase reaction D. de Sanctis et al. 4568 FEBS Journal 277 (2010) 4562–4574 ª 2010 The Authors Journal compilation ª 2010 FEBS mutant crystallized without arabinohexose, no Tris molecule is present in the catalytic site (data not shown). In the D171A BsArb43B mutant, additional electron density was observed on the other side of the binding cleft, opposite to where the arabinotriose molecule was modelled (Fig. 4B). This residual electron density could not be accounted for protein atoms or water molecules, but its paucity did not enable any addi- tional saccharide molecule to be inserted into the model. Nevertheless, this electron density could result from a partially occupied alternative binding of the polysaccharide chain or from a saccharide product that resulted from residual activity of the mutant. The C-terminal domain The most obvious structural difference between BsArb43B and other arabinanases is the presence of the additional C-terminal domain. This domain com- prises 103 amino acids (residues 367–470) organized in a short piece of a-helix and eight b-strands, arranged in a distorted b-barrel-like configuration (Fig. 1). Although the reported structures of arabinanases do not show such domains, other members of GH43 fam- ily contain an extra domain, the carbohydrate- binding module (CBM), namely b-xylosidase from G. staero- thermophilus (XynB3, Protein Data Bank code 2EXH) [27] and arabinoxylan arabinofuranohydrolase AXH- m2,3 from B. subtilis (BsAXH-m2,3; Protein Data Bank code 3C7E) [29]. Structural comparison of the BsArb43B C-terminal domain with these two struc- tures reveals that its orientation is different. In BsArb43B, the extra C-terminal domain is close to bla- de V of the catalytic domain, but the CBM domains in XynB3 and BsAXH-m2,3 are located close to blade I. In addition, the CBMs of XynB3 and BsAXH-m2,3 appear to be completely independent from the catalytic domain, interacting mainly through polar contacts. In BsArb43B, the putative CBM appears to interact more tightly with the N-terminal domain, and hydrogen bonds are observed between these two entities [residues Arg366(NH1)–Glu33(OE2), Arg366(NH2)–Glu33(OE1), Asp392(OD1)–Lys296(NZ), Asp392(O)–Arg255(NH2), Asp392(N)–Glu291(OE2), Lys398(O)–Thr302(N), Gln 443(OE1)–Tyr365(N)], together with a hydrophobic core nestled between these entities (residues Val256, Ala269, Val295, Met298, Tyr301, Trp359, Pro364, Ile387, Leu458 and Trp466). Structural alignment using DALI [30] or SSM from EBI [31] does not show any relevant matches between this domain and other structures in the databases. The highest hit obtained with both servers is the exclusion domain of dipeptidyl peptidase I or cathepsin C (Protein Data Bank code 1K3B [32]). The role of the C-terminal domain was further investigated by con- struction of two truncated versions of the enzyme. Based on structural and sequence alignments, two truncated proteins were engineered that lack 119 residues (trunc1) or 106 residues (trunc2) at the C- terminus (Fig. S2). The genes encoding the two mutant proteins were individually expressed in E. coli, but the proteins were not detected. The lack of accumulation in vivo indicates poor stability, and suggests that the presence of the C-terminal domain is crucial for the acquisition of the correct enzyme fold. Analysis of the interactions between the two domains led us to presume that expression of the truncated version of BsArb43B may expose the hydrophobic core described above and reduce the protein stability. An extra domain is found in putative arabinanases that are present in the genomes of other bacteria, in particular those of the Bacillus ⁄ Clostridium group. CBMs are commonly found by glycoside hydrolases, which utilize an insoluble substrate in order to attack this polysaccharide more efficiently. For this reason, CBMs retain the ability to concentrate enzymes onto the polysaccharide substrate, leading to more rapid degradation of the polysaccharide [33]. Detailed analy- sis of the putative CBM of BsArb43B does not show any evidence of extra sugar binding sites. However, determination of the surface charge distribution using adaptive Poisson-Boltzmann solver (APBS) [34] (Fig. 6) showed that the C-terminal domain presents a Fig. 6. Surface charge distribution of Abn2, viewed from the oppo- site side of the active site. It is possible to identify the entrance of the funnel between the blades of the propeller, which is partially negatively charged. The putative CBM shows a mostly positive charge. D. de Sanctis et al. Role of calcium in the glycosidase reaction FEBS Journal 277 (2010) 4562–4574 ª 2010 The Authors Journal compilation ª 2010 FEBS 4569 largely positively charged surface, due to the residues Lys376, Lys378, Lys421, Lys424, Arg442, Arg448 and Lys469, and this feature could be related to its hypo- thetical function. Whether this extra domain consti- tutes a CBM or is an evolutionary relic of a longer ancestoral enzyme is currently under investigation. It is worth noting that no CBM with a similar b-barrel-like shape has yet been reported. The b-trefoil folding of a protein of the CBM family 13 (http://www.cazy.org/) [13] resembles a b-barrel fold, but structural compari- son between the putative CBM of BsArb43B and CBM13 does not suggest any relevant similarity. Concluding remarks The work presented here shows that BsArb43B has a 3D fold that is different from those of other arabinan- ases with a known structure. In addition to the catalytic domain that is common to the other arab- inanases, the BsArb43B 3D fold comprises an extra C-terminal domain. Whether this extra domain is a CBM or has a different function is still under investi- gation. Detailed analysis of the binding cleft of BsArb43B and the other structurally determined arab- inanases showed that the exo-ABN from C. japonicus has a long loop that occludes one of the sides of the cleft, whereas all the endo-ABNs have loops of smaller and similar size that leave the binding cleft open at both sides, allowing it to act in endo mode. The pres- ent work also enabled precise identification of the metal in the active cleft as calcium, and suggested the nature of its role in the enzymatic mechanism. Based on data reported here, calcium appears to be impor- tant for the enzymatic mechanism of the enzyme, probably by directly influencing the protonation state of the catalytic carboxylate. In addition, these data also show that the histidine residue (His318) that coor- dinates with the calcium also plays a role in the enzyme mechanism by binding and stabilizing the substrate in the active site. Experimental procedures Substrates Debranched arabinan (linear a-1,5-l-arabinan, purity 95%) and a-1,5-l-arabinooligosaccharides (arabinohexose, purity 95%) were purchased from Megazyme International (Bray, Ireland). Bacterial strains and growth conditions Escherichia coli DH5a (Gibco BRL Richmond, CA, USA) was used for routine molecular cloning, and E. coli BL21 (DE3) pLysS [35] was used as the host for expression of the recombinant protein BsArb43B from Bacillus subtilis 168T + and mutant proteins. All strains were grown on Luria–Bertani medium [36], with kanamycin (30 lgÆmL )1 ), chloramphenicol (20 lgÆmL )1 ) and isopropyl-b-d-thiogalac- topyranoside) being added as appropriate. DNA manipulation and mutagenesis DNA manipulations were performed as described previ- ously [37]. PCR amplifications were performed in a MyCy- clerÔ thermal cycler (Bio-Rad, Hercules, CA, USA). A QIAprep Spin Miniprep kit (Qiagen, Valencia, CA, USA) was used to purify the plasmids. DNA was sequenced using an ABI PRIS BigDye Terminator Ready Reaction Cycle Table 3. Data collection statistics. BsArb43B D171A BsArB43B H318A Native BsArB43B b Beam line at European Synchrotron Radiation Facility ID23-1 ID14-3 ID29 Wavelength (A ˚ ) 1.06725 0.93100 1.0332 Detector ADSC Quantum Q315r ADSC Quantum 4 ADSC Quantum Q315r Distance 172.37 172.36 265.56 Resolution (A ˚ ) 1.50 1.79 1.90 Space group P1 P1 P1 Cell parameters a, b, c (A ˚ ) 51.9, 57.9, 85.6 51.8, 57.4, 85.5 51.9, 57.6, 86.2 a, b, c (°) 96.2, 91.8, 117.3 82.1, 88.2, 63.6 82.3, 87.9, 63.6 Number of unique hkl a 133 640 76 283 67 116 Completeness (%) a 94.9 (92.7) 93.3 (72.2) 95.7 (87.1) Mean I, r(I) a 13.5, 2.2 13.7, 3.3 14.5, 6.4 R symm a 0.038 (0.377) 0.077 (0.385) 0.047 (0.136) Multiplicity a 2.0 (2.0) 3.9 (3.4) 2.0 (1.9) a For the highest-resolution shell: 1.58–1.50 A ˚ for D171A, 1.89–1.79 A ˚ for H318A. b Data adapted from de Sanctis et al. 2008 [18]. Role of calcium in the glycosidase reaction D. de Sanctis et al. 4570 FEBS Journal 277 (2010) 4562–4574 ª 2010 The Authors Journal compilation ª 2010 FEBS Sequencing kit (Applied Biosystems, Foster City, CA, USA). Amino acid substitutions in BsArb43B were created using the QuikChange site-directed method (Strategene, La Jolla, CA, USA) using the respective mutagenic oligonu- cleotide pairs (Table S2) and plasmid pZI39 as template. Truncation of BsArb43B was performed based on sequence (first 363 residues) and structure (first 350 residues) align- ments. Briefly, the truncated C-terminus of the protein was amplified by PCR using primers ARA246 and ARA404 (sequence alignment) or primers ARA246 and ARA384 (structure alignment), with pZI39 as template (Table S2). The resulting 623 or 581bp DNA fragments, respectively, were digested using SalI and XhoI, and cloned into the same sites of pZI39. The presence of mutations and correct truncation of BsArb43B were verified by sequencing of the resulting plasmids. Protein expression and purification For protein over-production, E. coli BL21 (DE3) pLysS cells carrying the desired plasmid were grown on LB med- ium, and the extracellular BsArb43B and derived mutants were extracted from the periplasmic protein fraction by cold osmotic shock, as previously described [18]. Bio- chemical analyses revealed that all of the mutants were suc- cessfully expressed and had a migration pattern on SDS ⁄ PAGE identical to that of wild-type BsArb43B, except for the truncated versions, which were not detected. For purification of recombinant BsArb43B and the BsArb43B H318A and H318Q mutants, the periplasmic protein fraction was filtered and loaded onto a 1 mL Histrap column (Amersham Pharmacia Biotech, Piscataway, NJ, USA). The bound proteins were eluted by discontinu- ous imidazole gradient, and fractions containing more than 95% pure protein were dialysed overnight against a dialysis buffer (1 · phosphate buffer, 10% glycerol), and then frozen in liquid nitrogen and kept at )80 °C until further use. Enzyme assays The source of the enzyme was the periplasmic protein fraction of E. coli cultures or purified arabinanases. The enzyme activity was determined as previously described [10]. The reducing sugar content after hydrolysis of the polysaccharides was determined by the Nelson–Somogyi method, with l-arabinose as standard [11]. One unit of activity was defined as the amount of enzyme that produces 1 lmol of arabinose equivalents per minute. The kinetic parameters (apparent K m and V max values) were determined from the Lineweaver–Burk plot at optimum pH and tem- perature using linear a-1,5-l-arabinan as the substrate at concentrations ranging from 1 to 10 mgÆmL )1 . Thermal shift assays Samples were prepared by adding 5 · Sypro Orange (Molecular Probes, Carlsbad, CA, USA) to a mixture con- taining the protein solution in a 96-well thin-wall PCR plate sealed with optical-quality sealing tape (Bio-Rad) and Table 4. Refinement statistics. BsArb43B D171A BsArb43B H318A Native BsArb43B Number of protein atoms 7167 7160 7118 Number of solvent atoms 691 491 593 Number of hetero atoms 63 23 152 Number of sugar atoms 56 0 0 Final R factor (%) 14.73 15.84 14.07 Final free R factor (%) a 16.66 18.85 17.96 Mean B values (A ˚ 2 ) Wilson B 11.7 13.4 11.5 Protein 14.7 16.9 19.2 Solvent 24.1 (water) 23.3 (water) 26.5 (water) Overall 6.3 (Ca 2+ ) 17.5 (AHR) 9.5 (Ca 2+ ) 21.3 (TRIS) 12.9 (Ca 2+ ) 14.1 (TRIS) Maximal estimated error (A ˚ ) 0.044 0.071 0.071 Distance deviations Bond distances (A ˚ ) 0.017 0.010 0.015 Bond angles (A ˚ ) 1.403 1.185 1.471 Planar groups (A ˚ ) 0.006 0.005 0.005 Chiral volume deviation (A ˚ 3 ) 0.126 0.089 0.089 Ramachandran analysis (%) [44] Favourable 98.1 (924 ⁄ 942) 98.1 (925 ⁄ 943) 98.1 (909 ⁄ 927) Allowed 100 (942 ⁄ 942) 100.0 (943 ⁄ 943) 100.0 (927 ⁄ 927) a Calculated with 5% of reflections excluded from refinement. D. de Sanctis et al. Role of calcium in the glycosidase reaction FEBS Journal 277 (2010) 4562–4574 ª 2010 The Authors Journal compilation ª 2010 FEBS 4571 [...]... parameters for all the structures presented TLS thermal anisotropic parameterization was included in the final stages of refinement Each molecule was divided into two TLS groups, corresponding to the N- and C-terminal domains The structures of the mutant proteins H318A and D171A were solved by the molecular replacement method, using the program PHASER [41], with the model of the native protein as refined from the. .. datasets were integrated using MOSFLM [39] and scaled using SCALA [40] Data collection statistics for the triclinic crystal form of the mutant proteins are given in Table 3 Structure determination and refinement The structure of native BsArb43B was solved by the SAD method using Se-Met derivatives and data collected from a crystal in the orthorhombic form An initial structural model was obtained as previously.. .Role of calcium in the glycosidase reaction D de Sanctis et al heated in an iCycler iQ5 Real Time PCR detection system (Bio-Rad) from 20 to 90 °C, using increments of 1 °CÆ min)1 The fluorescence intensity change was measured using a CCD camera with excitation at 490 nm and emission at 530 nm The midpoint temperature of the protein unfolding transition, Tm, was calculated... (1968) Conformation of polypeptides and proteins Adv Protein Chem 23, 283–437 4574 Supporting information The following supplementary material is available: Fig S1 X-ray fluorescence spectra of a native Abn2 crystal Fig S2 Sequence alignment of the C-terminus of BsArb43B and other a-l -arabinanases Fig S3 Thermal shift essay for native BsArb43 and BsArb43 incubated with EDTA or EGTA Table S1 Residues involved... reductase and analysis of factors contributing to b-propeller folds J Mol Biol 269, 440–455 Neer EJ & Smith TF (1996) G protein heterodimers: new structures propel new questions Cell 84, 175–178 Role of calcium in the glycosidase reaction 25 Pons T, Naumoff DG, Martinez-Fleites C & Hernandez L (2004) Three acidic residues are at the active site of a b-propeller architecture in glycoside hydrolase families... refined from the triclinic crystal form as the search model, without the solvent molecules The structures were refined according to the procedure used for the native BsArb43B in the triclinic form; refinement statistics for all three structures (native BsArb43B, H318A and D171A mutant proteins) are reported in Table 4 Acknowledgements ´ Claudio M Soares Andrea Spallarossa (Department of Pharmaceutical Science,... Residues involved in the interface between the N- and C-terminal domains Table S2 Plasmids and oligonucleotides used in this study This supplementary material can be found in the online version of this article Please note: As a service to our authors and readers, this journal provides supporting information supplied by the authors Such materials are peer-reviewed and may be re-organized for online delivery,... Lebioda L & Minor W (2008) Data mining of metal ion environments present in protein structures J Inorg Biochem 102, 1765–1776 29 Vandermarliere E, Bourgois TM, Winn MD, van Campenhout S, Volckaert G, Delcour JA, Strelkov SV, Rabijns A & Courtin CM (2009) Structural analysis of a glycoside hydrolase family 43 arabinoxylan arabinofuranohydrolase in complex with xylotetraose reveals a different binding mechanism... different binding mechanism compared with other members of the same family Biochem J 418, 39–47 30 Holm L, Kaariainen S, Wilton C & Plewczynski D (2006) Using Dali for structural comparison of proteins Curr Protoc Bioinformatics 14, 5.1–5.5.24 31 Krissinel E & Henrick K (2004) Secondary-structure matching (SSM), a new tool for fast protein structure alignment in three dimensions Acta Crystallogr D Biol... (2001) Structure of human dipeptidyl peptidase I (cathepsin C): exclusion domain added to an endopeptidase framework creates the machine for activation of granular serine proteases EMBO J 20, 6570–6582 33 Bolam DN, Ciruela A, McQueen-Mason S, Simpson P, Williamson MP, Rixon JE, Boraston A, Hazlewood GP & Gilbert HJ (1998) Pseudomonas cellulose-binding domains mediate their effects by increasing enzyme substrate . of BsArb43B. Furthermore, to determine whether the role of the calcium is structural, thermal shift assays were per- formed to determine the T m of the protein in the. New evidence for the role of calcium in the glycosidase reaction of GH43 arabinanases Daniele de Sanctis 1,2, *, Jose ´ M. Ina ´ cio 1, * , ,