Cellvibrio japonicus alpha-L-arabinanase 43A has a novel five-blade beta-propeller fold

© 2002 Nature Publishing Group http://www.nature.com/naturestructuralbiology

letters (http://afmb.cnrs-mrs.fr/~cazy/CAZY/index.html)4. The α-1,5arabinanase Arb43A from Cellvibrio japonicus (formally known as Pseudomonas cellulosa)5 hydrolyzes linear arabinans almost exclusively into arabinotriose (A3), implying a chain-end exo activity, but the enzyme is also able to cleave in an endo fashion. Together this suggests a ‘processive’ mode of action in which, following an internal cleavage event, the enzyme proceeds along the polymer performing numerous catalytic events to liberate trisaccharides without releasing the polymer chain3. Here we Didier Nurizzo1, Johan P. Turkenburg1, 1,2 1 present the three-dimensional structure of Arb43A. The enzyme Simon J. Charnock , Shirley M. Roberts , Eleanor J. Dodson1, Vincent A. McKie3, Edward J. Taylor3, possesses a ‘non-velcroed’ five-bladed β-propeller; the topography at the active site provides insight into the unusual catHarry J. Gilbert3 and Gideon J. Davies1 alytic properties of this glycoside hydrolase.

Cellvibrio japonicus -L-arabinanase 43A has a novel five-blade -propeller fold

1

Structural Biology Laboratory, Department of Chemistry, University of York, Heslington, York YO10 5DD, UK. 2Present address: School of Applied and Molecular Sciences, Northumbria University, The Ellison Building, Newcastle upon Tyne NE1 8ST, UK. 3Department of Biological and Nutritional Sciences, University of Newcastle upon Tyne, Newcastle upon Tyne NE1 7RU, UK.

Published online: 19 August 2002, doi:10.1038/nsb835

Cellvibrio japonicus arabinanase Arb43A hydrolyzes the -1,5-linked L-arabinofuranoside backbone of plant cell wall arabinans. The three-dimensional structure of Arb43A, determined at 1.9 Å resolution, reveals a five-bladed -propeller fold. Arb43A is the first enzyme known to display this topology. A long V-shaped surface groove, partially enclosed at one end, forms a single extended substrate-binding surface across the face of the propeller. Three carboxylates deep in the active site groove provide the general acid and base components for glycosidic bond hydrolysis with inversion of anomeric configuration. Arabinan is a plant cell wall polymer featuring a backbone of α-1,5-L-arabinose in the furanoside ring form, further decorated with α-1,3- and α-1,2-linked arabinofuranosides1. The backbone is hydrolyzed by arabinanases2,3, members of the glycoside hydrolase (GH) family 43 of the CAZy classification

a

Three-dimensional structure of Arb43A The native Arb43A structure was solved by noncrystallographic symmetry averaging, using data in both P6522 and P1 crystal forms (see Methods). Residues 30–347 (the first 29 residues being a cleaved signal peptide) were built into the averaged 1.9 Å electron density map. The monomer of Arb43A is composed of a single five-bladed β-propeller (Fig. 1a). The propeller has a cylindrical shape with diameter and height of ∼40 Å and is present as a homodimer both in solution (data not shown) and in the crystal (Fig. 1b). Crystallographic symmetry in the P6522 form generates this dimer, whereas the P1 form contains three noncrystallographic dimers identical to those observed in the hexagonal crystals. Comparison of the molecules in the different crystal forms gives r.m.s. values of 0.18–0.29 Å between equivalent Cα positions. The angle of ∼90° between the central axis of the propeller of the two monomers suggests that the two active centers are unlikely to act on the same polysaccharide chain, and we see no evidence for cooperativity between the dimer components. The fold of Arb43A shows no homology with any of the 40 GHs for which a three-dimensional structure is known. Indeed, together with tachylectin (described below), Arb43A forms only the second member of the five-bladed β-propeller family; therefore, we believe that it is the first enzyme described with this fold.

c

b

Fig. 1 Comparison of the three-dimensional structure of Arb43A and tachylectin. a, Three-dimensional structure of Arb43A in divergent stereo view. The color is ‘ramped’ from N (blue) to C (green) terminus. The central ion is shown as a sphere. b, The dimer of Arb43A. The orientation of the two monomers suggests that each molecule of the dimer acts on a different substrate chain. c, The structure of tachylectin with the five ligands shown in ball-and-stick representation.

nature structural biology • volume 9 number 9 • september 2002

665

letters O

OH O

α-1,5 linked substrate

© 2002 Nature Publishing Group http://www.nature.com/naturestructuralbiology

OH O OH O O

O

O

O

OH O C5

OH O OH

C1

HO

OH OH O

O

OH

Therefore, we do not believe that there is evidence for a common ancestor for these two proteins. Tachylectin binds N-acetylglucosamine and N-galactosamine at five sites (Fig. 1c), reflecting the DNW and IGXGGW repeated motifs (underlined residues are involved in ligand recognition), respectively, of the protein. Arb43A displays a different mode of substrate binding (see below).

OH

OH O

O

O

OH O

OH

OH

O

HO

C OH

β-L arabinosyl product

HO

Fig. 2 The reaction catalyzed by Arb43A. The hydrolysis of α-1,5-Larabinans occurs with inversion of anomeric configuration.

The Arb43A propeller is based upon a five-fold repeat of blades composed of four-stranded β-sheets. A loop is inserted in the fourth β-strand of blade 3, disrupting the strand between residue 265 and 277. These blades are arranged radially around the central axis and are strongly twisted. The axial cavity is filled with water molecules and a single ion, whose temperature factor and coordination distances of 2.5–2.7 Å lead us to model it as a single occupancy chloride with temperature factors 25–20 Å2, although Ca2+ also remains a possibility. The ion displays pentagonal bipyramidal coordination, with the two axial ligands donated from the NE2 of His 391 and a water molecule, with five water molecules lying in the equatorial plane. The β-sheets from the five blades pack face to face, with hydrophobic interactions and hydrogen bonds as observed for other β-propeller proteins6. Most β-propellers so far observed are ‘closed’ by the completion of the C-terminal four-stranded sheet through incorporation of a strand from the N-terminus, or vice versa, colloquially termed ‘molecular velcro’. This is believed to provide considerable stabilization to the fold7–10. Nonvelcroed propellers are rare, having primarily been described only for the seven-bladed prolyloligopeptidase from porcine muscle11, where the resultant flexibility is believed to facilitate substrate transfer. In Arb43A, residues 38–74 form the four strands in the first blade, and the entire fifth blade is formed by residues from the C-terminus; thus, no classi- a cal velcro is present. There are, however, three hydrogen bonds from the C-terminal residues 344–346 to the tip of strand 4 of the first propeller blade, which may provide some stabilization to the fold (Fig 1a). The only other structure displaying the five-fold β-propeller is tachylectin-2 (ref. 9), although related fivemodule β/α propellers have recently been described12. Although the 236 amino acids of tachylectin display the same topology as the 347 residues of Arb43A, superposition of these proteins gives r.m.s values >8 Å.

Substrate-binding and catalysis by Arb43A Catalysis by members of GH-43 occurs with inversion of anomeric configuration13 (Fig. 2). By analogy with enzymes acting on pyranosides14, we expect the mechanism to involve a single displacement in which a catalytic acid group provides protonic assistance to the departure of the leaving group and a second carboxylate provides Brønsted base catalytic assistance, activating a water molecule for nucleophilic attack. To determine the structure of substrate in complex with Arb43A, an inactive mutant was generated by systematically mutating the cluster of three invariant carboxylates at the putative catalytic center, Asp 38, Asp 158 and Glu 221. Mutation of any of these three amino acids substantially reduced kcat and, in the case of D38E, there was a significant increase in KM, reflecting a reduction in affinity. The F114A mutant (described below) also displayed significantly reduced affinity as evidenced by a high KM (Table 1). Arb43A possesses six sugar-binding subsites, with cleavage of arabinohexaose (A6) occurring between sugars 3 and 4 (ref. 3). Crystals of the D158A mutant in the presence of A3 reveal apparent connected density for a hexasaccharide (Fig. 3a), although low occupancy and weak electron density between sugars three and four suggest that this density is probably an overlap of arabinotriosides. The pseudosymmetry of arabinofuranosides makes determining their orientation difficult because oligosaccharide chains built in either direction fit equally well into density. The sole criterion used to delineate orientation is the observation of the hydroxymethyl C5-O5 group at the nonreducing end; therefore, interpretation must be treated with caution. Regardless of orientation, Glu 221 is adjacent to the glycosidic oxygen atom between sugars 3 and 4, the likely site of bond cleavage, and is presumably the catalytic acid. In the modeled orientation of the substrate, Asp 38 lies 6 Å from the

b

Fig. 3 Three-dimensional structure of Arb43A in complex with substrate. a, Observed electron density for the arabinotriose complex of D158A Arb43A (maximum-likelihood / σA-weighted25,28 Fo – Fc synthesis contoured at 1 σ in divergent stereo). The position of the three catalytic carboxylates and Phe 114 are shown for reference (the position for Asp 158 is that from the native enzyme structure). The ligand has been modeled as a single hexasaccharide; although, it most likely results from an overlap of trisaccharides. b, Surface representation of Arb43A in complex with arabinotriose. The ligand is green; the surfaces associated with Glu 221 and Asp 38, red; and the side chain of Phe 114, yellow. The loop insertion, absent in endo arabinanases, that causes a kink at the end of the binding cleft is colored as a blue surface.

666

nature structural biology • volume 9 number 9 • september 2002

© 2002 Nature Publishing Group http://www.nature.com/naturestructuralbiology

letters anomeric carbon of the sugar at the –1 sub- Table 1 Activity of wild type and mutants of Arb43A against linear arabinan (LA) site and may function as the catalytic base, and arabinohexaose (A6) activating a water for single-displacement Arb43A kcat1 (s–1) KM1 (mg ml–1) kcat / KM1 Relative activity attack at C1. Although Asp 158 is impor(s–1 ml–1 mg) tant in catalysis (Table 1), its function is LA3 A64 unclear. In the native structure Asp158 lies Wild type 17.1 ± 1.6 2.6 ± 0.14 6.58 1 1 3.6–4 Å from Glu 221 and, thus, may funcD38E 1.5 × 10–3 ± 1.1 × 10–4 11.6 ± 0.7 1.3 × 10–4 4.3 × 10–5 5.0 × 10–5 tion in pKa modulation and orienting the D38N 2.6 × 10–4 ± 3.6 × 10–6 3.1 ± 0.1 7.7 × 10–5 4.0 × 10–5 4.8 × 10–6 catalytic acid, as seen in other glycoside –2 –2 F114A 1.4 ± 0.12 11 ± 4.3 1.3 × 10 2.1× 10 7.7 ×10–4 hydrolase structures15. D38A –2 – – 3.5 × 10–6 1.3 × 10–6 Arb43A liberates trisaccharide products D158A – – – 9 × 10–7 7.1 × 10–7 following initial internal cleavage of arabi–6 D158N – – – 2.0 × 10 1.3 × 10–6 nan. Processivity has been observed in E221A5 – – –