Crystal structure of levansucrase from the Gram-negative bacterium Gluconacetobacter diazotrophicus

Abstract

The endophytic Gram-negative bacterium Gluconacetobacter diazotrophicus SRT4 secretes a constitutively expressed levansucrase (LsdA, EC 2.4.1.10), which converts sucrose into fructooligosaccharides and levan. The enzyme is included in GH (glycoside hydrolase) family 68 of the sequence-based classification of glycosidases. The three-dimensional structure of LsdA has been determined by X-ray crystallography at a resolution of 2.5 Å (1 Å=0.1 nm). The structure was solved by molecular replacement using the homologous Bacillus subtilis (Bs) levansucrase (Protein Data Bank accession code 1OYG) as a search model. LsdA displays a five-bladed β-propeller architecture, where the catalytic residues that are responsible for sucrose hydrolysis are perfectly superimposable with the equivalent residues of the Bs homologue. The comparison of both structures, the mutagenesis data and the analysis of GH68 family multiple sequences alignment show a strong conservation of the sucrose hydrolytic machinery among levansucrases and also a structural equivalence of the Bs levansucrase Ca²⁺-binding site to the LsdA Cys³³⁹–Cys³⁹⁵ disulphide bridge, suggesting similar fold-stabilizing roles. Despite the strong conservation of the sucrose-recognition site observed in LsdA, Bs levansucrase and GH32 family Thermotoga maritima invertase, structural differences appear around residues involved in the transfructosylation reaction.

INTRODUCTION

Fructans (β-D-fructofuranosyl polymers with a terminal D-glucosyl moiety) constitute the primary carbohydrate reserve in 15% of higher plants comprising approx. 40000 species and appear in a wide range of bacterial and fungal species [1]. In plants, fructans are known to be involved in the stabilization of cellular membranes conferring protection from water stress caused by drought or low temperatures [2]. The physiological roles of fructans in bacteria have been less described, and it has been postulated that fructan biosynthesis may be involved in a variety of processes, such as survival of bacteria in soil, phytopathogenesis and symbiosis [3]. Of emerging importance is the use of fructo-oligosaccharides as health-promoting ‘pre-biotics’, which act as food sources for beneficial bacteria, such as bifidobacteria.

The synthesis of fructans starts with a transfructosylation reaction in which a sucrose molecule plays the role of fructosyl donor and with a second molecule of sucrose the initial acceptor of the fructosyl moiety (Scheme 1). Following initiation, the extension, the type of linkage and the branching of the fructan chain varies according to the enzyme and organism. Fructan synthesis in plants is carried out by the action of two or more FTFs (fructosyltransferases) that are able to produce complex mixtures of fructans with a great variety of linkages of the fructosides [2]. In bacteria, fructan biosynthesis is somewhat simpler because only one multifunctional enzyme is involved. This enzyme is named levansucrase (EC 2.4.1.10) when the product is a fructan of the type levan (β-2,6-linked polyfructan) or inulosucrase (EC 2.4.1.9) when the fructan is of the type inulin (β-2,1-linked polyfructan). Levansucrase catalyses the transfer of the fructosyl residue, from sucrose, to a variety of acceptors including water (sucrose hydrolysis), glucose (exchange), sucrose (oligofructoside synthesis) and fructan (polymerase reaction). The attack of sucrose can occur via O1 to form 1-kestose (1-kestotriose, isokestose; the basis of β-2,1-linked inulins) or via O6 to form the β-2,6-linked fructo-oligosaccharide 6-kestose. These reactions all occur via a Ping Pong mechanism involving the formation of a covalent fructosyl–enzyme intermediate [4–6] (Scheme 1).

Scheme 1. Various reactions catalysed by levansucrase.

Within the sequence-derived classification of GHs (glycoside hydrolases) and transglycosidases [7,8], bacterial FTFs are classified in family GH68 (http://afmb.cnrs-mrs.fr/CAZY) and, recently, the first three-dimensional structure of an FTF was published [9]. FTFs have also been shown to be structurally similar to sucrose-6-phosphate hydrolases, invertases, fructanases and eukaryotic FTFs found in family GH32. Together GH68 and GH32 thus comprise the β-fructofuranosidase clan GH-J.

The recent crystal structure of Gram-positive bacterium Bacillus subtilis (hereafter Bs) levansucrase and its complex with sucrose revealed the structural determinants for substrate recognition and the spatial disposition of three key catalytic acidic residues at the active site. The structure displays a five-bladed β-propeller topology that is shared with the elucidated crystal structures of Thermotoga maritima invertase (GH32) [10], Cellvibrio japonicus α-L-arabinanase (GH43) [11] and the Cichorium intybus fructan 1-exohydrolase IIa (GH32) [12]. Despite their high degree of structural similarities in the disposition of catalytic residues at the active site, the GH43 α-L-arabinanase operates with inversion of the anomeric configuration of the substrate, while levansucrases and invertases are ‘retaining’ enzymes.

The endophytic Gram-negative bacterium Gluconacetobacter diazotrophicus (hereafter Gd) secretes a constitutively expressed levansucrase (LsdA), which is responsible for utilization of the natural substrate sucrose [5,13]. The kinetic parameters of the levansucrases from the Bs and Gd are of the same order of magnitude in the initial step of the reaction, the formation of the fructosyl–enzyme intermediate, but there are remarkable differences in the affinity of the intermediates for the various fructosyl acceptors [5,14]. Bs levansucrase catalyses the formation of high-molecular-mass levan without accumulation of transient oligofructans, whereas Gd levansucrase synthesizes and accumulates a large amount of β-2,1-linked tri- and tetra-saccharides, with a much lower yield of levan under the same reaction conditions [5]. The structural determinants for the polymerization reaction of FTFs remain at present poorly understood, not least because a single structure makes it hard to mine the 48 (as of 25 April 2005) GH68 family members. The combined analysis of protein sequences, mutagenesis data and three-dimensional structures of levansucrases with different catalytic performances should thus contribute to a deeper insight into the polymerization mechanism of FTFs. In the present paper, as part of this process, we report the crystal structure of Gd levansucrase at a resolution of 2.5 Å (1 Å=0.1 nm) in its ligand-free state and limited biochemical analyses to support a detailed comparison with its Bs homologue.

MATERIALS AND METHODS

Bacterial strains and media

Escherichia coli strains DH5α [Φ80dlacZΔM15 recA1 endA1 gyrA96 thi-1 hsdR17 (r_K⁻ m_K⁺) supE44 relA1 deoR Δ(lacZYA-argF)U169] and S17-1 {thi pro hsdR recA [RP4-2(Tc::Mu) (Km::Tn7)]} [15] were used as a cloning host and as a conjugative donor host respectively. E. coli strains were grown at 37 °C in Luria–Bertani broth supplemented with ampicillin (50 μg/ml) and tetracycline (12.5 μg/ml) when needed. Gd strains SRT4 (wildtype, Lev⁺) and AD5 (SRT4 lsdA::nptII-ble cassette, Lev⁻) [16] were grown at 28 °C in LGIE medium containing 5% (w/v) sucrose and 1% (v/v) glycerol [17]. Chloramphenicol (30 μg/ml) and tetracycline (20 μg/ml) were added to LGIE agar medium to select the transconjugants.

Site-directed mutagenesis

Plasmid pALS137 carries the 3.2-kb HindIII/NruI fragment from plasmid pALS5 [17] inserted into pUC8 digested with HindIII/SmaI. Plasmid pALS137 containing the levansucrase gene (lsdA) under its own promoter was used as template for mutagenesis using the GeneEditor Kit (Promega). The oligonucleotides 5′-GATGTCTGGGTCTGGAACACGTGGACCCTGATCGAC-3′, 5′-CCAGCGTGGCGTCGCGAACAGCACCGAGGCCGATC-3′ and 5′-CTGATTTCGGCCAACAGCGTCAATGATCAGACCGAACGGCCG-3′ were used to generate the mutations D135N, C339S and C395S respectively (amino acid substitutions are in bold, restriction sites created by silent mutations are underlined). Plasmids from E. coli transformants resistant to the GeneEditor Antibiotic Selection Mix were screened by digestion with PmlI (D135N), NruI (C339S) and BclI (C395S), and mutations were confirmed by DNA sequencing. The plasmids carrying the mutations D135N, C339S and C395S in the lsdA gene were named pALS191, pALS201 and pALS203 respectively.

Enzyme expression in lsdA::nptII-ble Gd mutant

Plasmids pALS137, pALS190, pALS201 and pALS203 were HindIII-linearized and inserted into the broad-range cloning vector pRK293 [18]. The resulting plasmids were named pALS148, pALS192, pALS205 and pALS207 respectively and were used to transform E. coli S17-1 (tra⁺). The conjugal mating procedure as described by Arrieta et al. [17] was used to mobilize plasmids pALS148, pALS192, pALS205 and pALS207 into the Gd mutant AD5, in which the lsdA gene is replaced by the nptII-ble cassette.

Crystallization, data collection and refinement

Details of protein expression, purification and crystallization were as reported previously [19]. Crystals grew over 2 weeks using ammonium sulphate in the range 0.35–0.6 M in 0.1 M sodium citrate, pH 5.6, with 30–60% ethanol and the detergents SDS (1.5–5.0 mM) and β-octylglucoside (10–40 mM) present. Handling crystals grown in ∼50% ethanol was extremely difficult and approx. 50 crystals had to be screened in order to gain a single ordered specimen for data collection. A 2.5 Å resolution dataset was collected from a single crystal at 100 K on beamline ID14-EH1 [ESRF (European Synchrotron Radiation Facility), Grenoble, France] using an ADSC Quantum-4 detector. Diffraction images were integrated with MOSFLM and scaled and reduced with SCALA from the CCP4 suite of programs [20]. Data collection statistics are summarized in Table 1. The structure was solved by molecular replacement using the co-ordinates of Bs levansucrase as a search model (Protein Data Bank accession code 1OYG) with all residues (except alanine and glycine) truncated to serine. The program AMoRe [21] was used with data in the resolution range 20–4 Å and an outer radius of Patterson integration of 37 Å.

Table 1. Summary of data collection and refinement statistics.

Values in parentheses refer to the highest resolution shell, 2.64–2.5 Å, for the final dataset. R_sym=Σ(I−〈I〉)²/ΣI².

Parameter	Value
Data collection
High resolution (Å)	2.5
Completeness (%)	99.2 (98.5)
Multiplicity	3.3 (3.2)
Rsym	0.078 (0.318)
〈I/σI〉	11.6 (4.6)
Refinement
Resolution used in refinement (Å)	105.4–2.5
Total number of non-hydrogen atoms	7848
Number of water molecules	194
Number of sulphate ions	4
Rcryst (%)	19.6
Rfree (%)*	24.9
Average B-factor
Protein atoms (Å2)	25.7
Sulphate ions atoms (Å2)	74.3
Solvent atoms (Å2)	34.1
R.M.S.D. from ideal values†
Bond lengths (Å)	0.014
Bond angles (°)	1.51
B-factor for bonded atoms
Main chain bonds (Å2)	0.612
Main chain angles (Å2)	0.786
Side chain bonds (Å2)	1.232
Side chain angles (Å2)	1.920
Protein Data Bank accession code	1W18

* R_free calculated using 5% of total reflections omitted from refinement.

† Data from [45].

For refinement, 5% of the observations were set aside for cross-validation analysis [22] and were used to monitor various refinement strategies. Simulated annealing by slow cooling and positional refinements were initially performed on the molecular replacement solution with CNS [23] using NCS (non-crystallographic symmetry) restraints. CNS density modification procedures and 2-fold map averaging were applied to the calculated 2Fo-Fc maps, and this allowed the placement of the structurally most conserved residues. After several cycles of NCS refinement with NCSref [20] interspersed with cycles of manual building with Xfit [24], it was possible to place approx. 70% of the residues. Using phases calculated from this incomplete model, ARP/wARP [20] built approx. 90% of the residues as serine, valine, alanine and glycine that were later mutated to the correct side chains. The rest of the residues, mainly in loop regions, were placed manually with Xfit [24]. Final refinement cycles were performed with REFMAC [25] using NCS restraints and TLS (translation, libration and screw-rotation) corrections [26]. Water molecules and four sulphate ions were placed at peak positions of σ_a-weighted [27] Fo-Fc maps and 2Fo-Fc maps greater than 3σ and 1σ respectively and were checked at the graphic display. A summary of refinement statistics is shown in Table 1. The final model was validated against composite annealed omit maps from CNS. Ramachandran [28] statistics produced by PROCHECK [29] indicate that 82.9% of the residues are located in the most favoured regions of the plot and 16.1% in additional allowed regions. Eight residues were found in generous allowed regions with no residues located in the disallowed regions. Co-ordinates have been deposited at the Protein Data Bank [30] under accession code 1W18.

RESULTS AND DISCUSSION

Structure of Gd levansucrase LsdA

The open reading frame of Gd levansucrase (LsdA) encodes a 584-residue protein with a 30-residue N-terminal signal peptide that is cleaved off during secretion. The mature protein thus comprises a 553-residue polypeptide (numbering 31–584) with an N-terminal pyroglutamate. LsdA contains three cysteine residues (Cys¹⁵⁷, Cys³³⁹ and Cys³⁹⁵), with Cys³³⁹ and Cys³⁹⁵ engaged in a disulphide bridge [31].

The crystal structure of LsdA (residues 62–554) was solved by molecular replacement using the homologous Bs levansucrase [9] as the search model at a resolution of 2.5 Å and was subsequently refined to crystallographic R-factors of 19.6% (R_cryst) and 24.9% (R_free) (Table 1). The asymmetric unit contains two copies of the polypeptide chain, and, in addition, 194 water molecules and four sulphate ions have been modelled. Tight NCS restraints were maintained throughout the refinement and model-building harnessed electron density maps averaged according to the co-ordinate-derived NCS operators.

LsdA displays the five-bladed β-propeller fold (Figure 1), as first reported for the N-acetylglucosamine-binding tachylectin [32] and as first observed as an enzyme fold for the GH43 arabinanase [11] and then subsequently for the invertase from T. maritima [10] and the fructan 1-exohydrolase from Ci. intybus [12]. The five β-sheet modules are arranged in consecutive order with 5-fold pseudosymmetry around a central cavity. As in other β-propellers described so far, the four β-strands composing each β-sheet display a strong twist with an average of 90° between the first and the last strand. Similarly to Bs levansucrase and arabinanase, LsdA does not possess ‘molecular Velcro’, a term colloquially used to describe the closure of the propeller fold in which the N-terminal first strand forms the outermost β-strand of the C-terminal blade [33,34]. Fold stabilization is, however, provided by the packing of the N- and C-termini of the polypeptide. As in Bs levansucrase, the LsdA N-terminus runs along the perimeter of blade IV forming a clamp-like loop that adds a fifth β-strand to blade III (numbered as βIIIE in Figure 2). Additionally in LsdA, the N-terminus (residues 76–78) forms a short two-stranded antiparallel β-sheet that extends to a β-strand (residues 442–444) located in the insertion between βIVC and βIVD (β-strands numbered as β1 and β4). Different from Bs levansucrase, the C-terminus in LsdA does engage in hydrogen-bonding interactions to the central core through the formation of a two-stranded parallel β-sheet between β-strand 2 (residues 147–150) and β-strand 5 (residues 542–545).

Figure 1. Three-dimensional structure of LsdA.

Superior (a) and lateral (b) stereo views of the five-bladed β-propeller fold. The colour is ‘ramped’ from N- (blue) to C- (red) terminus. Catalytic residues Asp¹³⁵, Asp³⁰⁹ and Glu⁴⁰¹ are shown in ball-and-stick representation. (c) Stereo view of the electron density map (contoured at 1σ level) ‘carved’ around catalytic residues and other residues involved in the hydrogen-bond (broken lines) network at the active site. These Figures were prepared with PYMOL [47].

Figure 2. LsdA (Gd_lvs) and Bs levansucrase (Bs_lvs) sequence alignment based on structural superimposition generated by TOP [20].

Secondary-structural elements were assigned using DSSP [48] and are indicated by arrows (β-strands) and squiggles (helices). These elements were labelled following the Bs levansucrase numbering in [9]. A roman numeral (I–V) is assigned to each β-sheet module, and, in each module, the β-strands are numbered with a capital letter (A–D). α- and 3₁₀-helices (labelled α and η respectively) are numbered consecutively. Strictly conserved residues are boxed. The Figure was produced with ESPript [49].

Comparison with the Bs levansucrase

LsdA presents an overall sequence identity with its Bs homologue of 26%, but, in the 283 structurally equivalent residues, the sequences share 31% identity. The secondary-structure elements superimpose well, with an R.M.S.D. (root mean square deviation) of 1.24 Å over 283 equivalent Cα atoms. Not surprisingly, greater variation is observed in insertions into surface loops, a feature commonly found in the β-propeller fold [35]. Several helices, including short stretches of 3₁₀-helix, are conserved along the sequence alignment (Figure 2). In both structures, the longest insertion is found between β-strands IIIB and IIIC, where three helical stretches are interspersed.

LsdA contains a disulphide bridge (Cys³³⁹–Cys³⁹⁵) that links the extended loop between β-strands IIIB and IIIC with the insertion located between blades III and IV. The replacement of either Cys³³⁹ or Cys³⁹⁵ by serine reduced the k_cat for sucrose hydrolysis approx. 60-fold (Table 2). Although equivalent cysteine residues appear in the levansucrases from the high-GC content Gram-positive species Arthrobacter sp. K-1 (GenBank® accession number BAB72022) and Actinomyces naeslundii (GenBank® accession number AAG09737) and the Gram-negative species Burkholderia cepacia R1808 (GenBank® accession number ZP_00218900) and Burkholderia pseudomallei K96243 (GenBank® accession number YP_110564), there is no general conservation of this disulphide bridge among GH68 family members. Indeed, a pair of cysteine residues is absent in most Gram-positive FTFs, including Bs levansucrase. Levansucrases from Gram-negative bacteria Rahnella aquatilis, Pseudomonas syringae and Pseudomonas aurantica display two highly conserved cysteine residues elsewhere, which are not equivalent to those found in LsdA and that might also be implicated in a disulphide bridge. It has been reported that the incubation with dithiothreitol or 2-mercaptoethanol decreased the enzyme activity of Ps. syringae levansucrase [36]. In Zymomonas mobilis levansucrase the substitution of serine for cysteine residues 121, 151 and 244 abolished the levan-forming activity and halved the sucrose hydrolysis activity [37]. In contrast, Bs levansucrase contains a Ca²⁺ ion connecting residues Asp²⁴¹, Gln²⁷², Leu³⁰⁸ and Asn³¹⁰ located in the longest insert between β-strands IIIB and IIIC, with Asp³³⁹ placed in the insertion between blades III and IV. Leu³⁰⁸ is not found strictly conserved in Gram-positive levansucrases, possibly due to the fact that the Ca²⁺ interaction with this residue is through the backbone carbonyl oxygen. The residues involved in Ca²⁺ binding are strictly conserved among FTFs from Gram-positive bacteria, except the high-GC content species Ac. naeslundii and Arthrobacter sp., but variable in the levansucrases of Gram-negative bacteria (see partial alignment in Figure 3). Thus the activity of LsdA and other levansucrases from Gram-negative bacteria is not affected in the presence of metal ions, such as Ca²⁺, or the chelating agent EDTA [5,37,38]. In contrast, FTFs from the Gram-positive species Streptococcus mutans [39], Streptococcus salivarius [6] and Lactobacillus reuteri [40,41] show maximal activity only upon addition of 1 mM Ca²⁺. The fact that both the disulphide bridge in LsdA and the Ca²⁺-binding site in Bs levansucrase map in structurally equivalent regions in their respective molecular architectures suggests that these features play a similar role in structural maintenance of the catalytic glutamate residue (Glu⁴⁰¹ in LsdA and Glu³⁴² in Bs levansucrase), found in the highly conserved ‘D(E/Q)(T/I/V)ER’ motif, poised for catalysis [9].

Table 2. Kinetic parameters of wild-type and site-directed mutant enzymes.

The native and mutant enzymes were purified from culture supernatants as described by Arrieta et al. [17] and quantified by the Bradford procedure using BSA as a standard. The reaction mixtures (50 μl) containing 200 mM sucrose in 0.1 M sodium acetate, pH 5.0, were incubated with 0.3 unit of purified enzyme at 30 °C. After 15 min, the reaction was stopped by heating in a boiling-water bath for 5 min. Glucose released from sucrose hydrolysis was measured by a glucose oxidase/peroxidase-coupled colorimetric reaction. The k_cat was calculated using 60000 as the M_r of a functional unit of levansucrase. The K_m for sucrose was determined within a substrate concentration range 1–150 mM at 30 °C and pH 5.0. The kinetic parameters were determined by non-linear regression analysis of substrate–velocity plots using the LEONORA program [46]. One unit of enzyme activity is defined as the amount of enzyme releasing 1 μmol of glucose/min based on initial-velocity measurements under the following conditions: 0.2 M sucrose in 0.1 M sodium acetate buffer, pH 5.0, at 30 °C. Values are means±S.D. (n=6).

Strain	Mutation	kcat (min−1)	Km (mM)	kcat/Km (M−1·min−1)
AD5 (pALS148)	Wild-type	(3.9±0.2)×103	11.9±0.4	3.3×105
AD5 (pALS192)	D135N	1.7±0.2	12.3±0.9	1.4×102
AD5 (pALS205)	C339S	66.9±2.3	13.8±1.2	4.8×103
AD5 (pALS207)	C395S	71.8±6.4	13.1±1.4	5.5×103

Figure 3. Multiple sequence alignment of selected fragments of GH68 family members.

The sequences are divided into two groups, Gram-negative and Gram-positive levansucrases headed by LsdA (Gd_lvs) and Bs levansucrase (Bs_lvs) respectively. Protein identifiers and access codes for each sequence in the SwissProt/TrEMBL and GenBank® databases are as follows: Gd LsdA (Gd_lvs, Q43998); B. pseudomallei K96243 levansucrase (SACB_Burps, YP_110564); B. cepacia R1808 hypothetical protein (hyp_Burce, ZP_00218900); Z. mobilis levansucrase (SACB_ZYMMO, Q60114, Q06487, Q60116), and invertase (invB_ZYMMO, AAA61488); Gluconacetobacter xylinus levansucrase (LsxA_Gxyl, BAA93720); Ps. syringae pv. tomato levansucrase (SACB_Psyr1, AAO54974; SACB_Psyr2, AAO55819; SACB_Psyr3, AAO59056); Ps. syringae pv. glycinea levansucrase (SACB_PSESG, O52408; SACB_PsyrG2, AAK49952); Ps. syringae pv. phaseolicola levansucrase (SACB_PSESH, O68609); R. aquatilis levansucrase (SACB_RAHAQ, O54435); Erwinia amylovora levansucrase (SACB_ERWAM, Q46654); Ps. aurantiaca levansucrase (SACB_Paura, AAL09386); Bs levansucrase (Bs_lvs, P05655, P70984); Bacillus stearothermophilus levansucrase (SACB_BACST, P94468); Bacillus amyloliquefaciens levansucrase (SACB_BACAM, P21130); Bacillus licheniformis unnamed protein product (unk_Blich, CAF05486); Clostridium acetobutylicum levansucrase (SACB_Cace, CAC1772; SACB_Cace1, CAC1774); Bacillus sp. V230 β-fructofuranosidase (bff_BspV230, BAA32083); Pa. polymyxa levansucrase (SACB_Ppol, CAB39327); Streptococcus salivarius levansucrase (SACB_STRSL, Q55242); L. reuteri inulosucrase (inu_Lreu, AAN05575), levansucrase (SACB_Lreu, AAO14618) and unnamed protein product (unk_Lreu, CAD19335); Lactobacillus johnsonii levansucrase (SACB_Ljohn, NP_964768); Streptococcus mutans levansucrase (SACB_STRMU, P11701); Lactobacillus sanfranciscensis levansucrase (SACB_Lsanf, CAD48195); Arthrobacter sp. K-1 β-fructofuranosidase (bff_AspK1, BAB72022); Ac. naeslundii FTF (ftf_Anae, AAG09737). Secondary-structural elements were assigned using DSSP [48] and are indicated by ‘e’ (β-strands) and ‘h’ (helices). Catalytic residues Asp¹³⁵, Asp³⁰⁹ and Glu⁴⁰¹ are boxed.

The active site of LsdA

Three strictly conserved acid residues have been identified in alignments of protein sequences corresponding to GH families 32, 43, 62 and 68 [42] (boxed residues in Figure 3). Site-directed mutagenesis of these carboxy residues has been performed for various representatives of GH68. On the basis of the mutagenesis and the structural information, it has been proposed that acidic residues Asp¹³⁵, Asp³⁰⁹ and Glu⁴⁰¹ of LsdA act as the catalytic nucleophile, the ‘transition-state stabilizer’ and the general acid–base catalyst respectively. Substitution of asparagine for Asp¹³⁵ resulted in an enzyme with substantially impaired activity, with a 2300-fold decrease in k_cat (Table 2). It has also been reported that mutation of Asp³⁰⁹ to asparagine in the RPD (Arg-Pro-Asp) motif reduced the k_cat 75-fold, again without change in the K_m [38].

In the crystal structure of LsdA, the catalytic constellation of carboxy residues is located at the bottom of a funnel-shaped cavity at the centre of the β-propeller (Figure 1). Each one of these residues, namely Asp¹³⁵, Asp³⁰⁹ and Glu⁴⁰¹, belong to the first and innermost β-strand of blades I, III and IV respectively. This feature is commonly found in this type of fold, where the surface of the central cavity is mostly lined by residues that belong to the first strand of each modular sheet [35]. The three carboxylic residues in LsdA are perfectly superimposable on to the equivalent residues in Bs levansucrase and are spaced some 4.8–7.4 Å from each other. The proposed catalytic nucleophile, Asp¹³⁵, side chain forms two hydrogen bonds with side chains of Ser²²⁵ and Ser⁴⁸⁵ (Figure 1c); the same pattern of interactions is present in Bs levansucrase for the equivalent residue. Both serine residues are found highly conserved in the GH68 family multiple sequence alignment, except that the equivalent to Ser²²⁵ in some of the Gram-negative bacteria, including Pseudomonas, Rhanella and Erwinia species, have an alanine residue in this position (Figure 3). Interestingly, the LsdA motif Tyr⁴⁸⁴-Ser⁴⁸⁵-(His/Tyr/Trp)⁴⁸⁶ is strictly conserved in GH68 family, except in Z. mobilis invertase and Paenibacillus polymyxa levansucrase, where Ser⁴⁸⁵ is found replaced by alanine. The structurally equivalent residues of LsdA Ser⁴⁸⁵ and Bs levansucrase Ser⁴¹² in T. maritima invertase and Ci. intybus fructan 1-exohydrolase are Ala²⁴¹ and Ala²⁷⁵ respectively, and, in the multiple sequence alignment of the GH32 family, an alanine residue is found highly conserved in this position (this motif was described as motif F in [43]). The high conservation of serine in this position in GH68 levansucrases and its replacement by alanine in GH32 invertases, levanases or sucrases suggests that this H-bond interaction could play an important role in the orientation of the nucleophilic residue during transfructosylation catalysis.

The side chain of Asp³⁰⁹ forms hydrogen bonds to Ser²²⁵, the Arg⁴⁰² backbone amide nitrogen and a water molecule that, at the same time, forms a hydrogen bond with the catalytic residue Glu⁴⁰¹. The Glu⁴⁰¹ side chain in turn establishes hydrogen bonds with side chains of Glu³²⁷ and Tyr⁴⁸⁴ and a salt bridge with Arg³⁰⁸ (Figure 1c). The equivalent positions of LsdA Arg³⁰⁸ and Tyr⁴⁸⁴ are found strictly conserved in GH68 family and are also conserved in GH32 family members (described as motifs D and F in [43]). The LsdA equivalent Glu³²⁷ position is strictly conserved in GH68 family, but is highly variable in representatives of GH32 family.

Analysis of the sucrose complex of the inactive mutant of Bs levansucrase (Protein Data Bank accession code 1PT2) [9] revealed the residues involved in substrate recognition. All of these residues, except for Gln³⁹⁹/Glu³⁴⁰ and His⁴¹⁹/Arg³⁶⁰ (LsdA/Bs levansucrase numbering), are present in LsdA. Indeed, structural overlap of this subset of co-ordinates gives a R.M.S.D. of 0.75 Å for the Cα atoms of residues that are responsible for sucrose recognition and catalysis. This close similarity allows the modelling and analysis of contacts of the sucrose molecule at the active site in LsdA. Figure 4 shows the LsdA sucrose-binding site and the potential hydrogen-bonding interactions of the sucrose molecule with LsdA based on the actual interactions with Bs levansucrase inactive mutant (Protein Data Bank accession number 1PT2). Most of the hydrogen-bonding interactions observed in the Bs levansucrase sucrose complex appear to be conserved in the modelled interaction of sucrose and LsdA. This high degree of structural conservation at the sucrose-binding site is in agreement with the enzymatic data that report kinetic constants of the same order of magnitude for the initial step of the transfructosylation reaction [5], i.e. the formation of the fructosyl–enzyme intermediate, and the overall sucrose hydrolysis reaction (Table 3). In addition, the residues involved in the hydrogen-bond network of interactions at the sucrose hydrolysis active site are found strictly conserved in the GH68 sequence alignment (Figure 3).

Figure 4. Sucrose-binding site of LsdA.

The Figure shows an overlap of the co-ordinates of the Bs levansucrase E342A-sucrose complex (Protein Data Bank accession code 1PT2) with the Gv levansucrase, LsdA. Potential hydrogen-bonding interactions are represented by broken lines. LsdA residues are bond-coloured green and Bs levansucrase are turquoise. Labels of depicted residues follow the same colour pattern.

Table 3. Apparent kinetic constants for the Bs and Gd levansucrases (adapted from [5]).

Values are given for the (forward) rate of intermediate formation (k₊₂) and subsequent hydrolysis, levan (lev) production and inulin (inu) production. Under conditions of 1 M sucrose, the Gd enzyme transfers approx. one-third of the fructosyl–enzyme intermediate to sucrose to form 1-kestose. Such a reaction was not observed for the Bs enzyme.

Origin	k+2 (s−1)	khydrolysis (s−1)	klev (M−1·s−1)	kinu (M−1·s−1)
Gd	550	51	1×104	Not detected
Bs	260	35	4×104	7×103

The sucrose hydrolytic machinery is located at the bottom of the catalytic pocket, and the residues that line this zone are located in the first β-strand of each blade. Immediately above this zone and surrounding the modelled glucosyl moiety of the sucrose molecule, it is possible to describe a fringe of residues that belong to the inserts that link β-strand B with C in each blade (see Figure 4). His⁴¹⁹ of LsdA is located in this region and is the structural equivalent of Bs levansucrase Arg³⁶⁰ and the sequence equivalent of Z. mobilis levansucrase His²⁹⁶. Indeed, in Gram-negative levansucrases, a histidine residue is found strictly conserved in the equivalent position of LsdA His⁴¹⁹, while in Gram-positive representatives an arginine residue is present (see the alignment in Figure 3). In both LsdA and Bs levansucrase, a tyrosine residue, whose aromatic ring is placed parallel to the plane of the histidine ring or the guanidinium group in LsdA His⁴¹⁹ or Bs levansucrase Arg³⁶⁰ (Tyr⁴⁸⁴ in LsdA and Tyr⁴¹¹ in Bs levansucrase) is found strictly conserved in all levansucrases. Both Bs levansucrase Arg³⁶⁰ and Z. mobilis levansucrase His²⁹⁶ are reported to be crucial for the catalysis of the transfructosylation reaction [14,44], suggesting that a positively charged and electron delocalizing environment is required around this position. This suggestion is supported further by the fact that T. maritima invertase [10] does not have a structural equivalent residue to LsdA His⁴¹⁹ or Bs levansucrase Arg³⁶⁰, and Ci. intybus fructan 1-exohydrolase shows a glutamate residue (Glu²³⁴) pointing in the opposite direction to Bs levansucrase Arg³⁶⁰ [12]. The presence of residual transfructosylation activity in Bs levansucrase Arg³³¹→Ser/Leu mutants [14] also indicates that the structural determinants for the transfructosylation catalysis are more complex and at present not fully understood. Indeed, the absence of structural data for complexes between fructosyl acceptors and levansucrases makes it difficult to reach conclusions about what determines the outcome of the transfructosylation reaction in which LsdA releases large amount of 1-kestotriose (Glc- and kestotetraose in the initial phase of the reaction [5]), while Bs levansucrase synthesizes long-chain fructans under similar reaction conditions (Table 3). The crystallographic structures currently available of members of GH68 and GH32 families, together with the sequences and biochemical data, strongly support the hypothesis that FTFs evolved from invertases through comparatively small mutational events [2]. The LsdA structure presented in the present paper sheds some light on some of these changes and should help inform subsequent mutagenesis experiments in order to explore and harness the structural determinants that are responsible for transfructosylation.