Characterization of the First α-(1→3) Branching Sucrases of the GH70 Family*

Leuconostoc citreum NRRL B-742 has been known for years to produce a highly α-(1→3)-branched dextran for which the synthesis had never been elucidated. In this work a gene coding for a putative α-transglucosylase of the GH70 family was identified in the reported genome of this bacteria and functionally characterized. From sucrose alone, the corresponding recombinant protein, named BRS-B, mainly catalyzed sucrose hydrolysis and leucrose synthesis. However, in the presence of sucrose and a dextran acceptor, the enzyme efficiently transferred the glucosyl residue from sucrose to linear α-(1→6) dextrans through the specific formation of α-(1→3) linkages. To date, BRS-B is the first reported α-(1→3) branching sucrase. Using a suitable sucrose/dextran ratio, a comb-like dextran with 50% of α-(1→3) branching was synthesized, suggesting that BRS-B is likely involved in the comb-like dextran produced by L. citreum NRRL B-742. In addition, data mining based on the search for specific sequence motifs allowed the identification of two genes putatively coding for branching sucrases in the genome of Leuconostoc fallax KCTC3537 and Lactobacillus kunkeei EFB6. Biochemical characterization of the corresponding recombinant enzymes confirmed their branching specificity, revealing that branching sucrases are not only found in L. citreum species. According to phylogenetic analyses, these enzymes are proposed to constitute a new subgroup of the GH70 family.

Numerous lactic acid bacteria from the Leuconostoc genus isolated from different habitats, such as sugar juice, fermenting vegetables, or dairy products, have long been known to produce slimes in sucrose solution (1,2). These slimy compounds were rapidly characterized as dextrans, homopolymers of ␣-D-glucopyranosyl units mainly linked by ␣-(136) linkages. Dextrans with a high content of ␣-(136) linkages and a low degree of branching found their first industrial applications in the 1940s as a source of synthetic blood volume expanders (3). These findings motivated Jeanes et al. (22) to investigate the diversity of polymers produced during sucrose fermentation by different strains of lactic acid bacteria. A total of 96 strains were screened, and their dextrans were structurally characterized. In this study one strain, Leuconostoc mesenteroides NRRL B-742, also known as L. mesenteroides ATCC 13146 and renamed Leuconostoc citreum NRRL B-742 (4), received particular attention. Indeed, this strain, first isolated from a can of spoiled tomatoes in 1927 (1), was shown to produce two types of dextrans that differed by their structures and degree of solubility from sucrose. The less soluble polymer contained 75% of ␣-(136) linkages and 15% of ␣-(134) linkages, whereas the more soluble polymer displayed an uncommon comb-like structure consisting of a linear backbone of ␣-(136)-D-glucopyranosyl residues grafted with one ␣-(133)-linked glucosyl unit on every glucosyl moiety (5)(6)(7)(8)(9). Concomitantly with those findings, dextrans were shown to be synthesized by ␣-transglucosylases, commonly named glucansucrases, and today classified in the family 70 of glycoside hydrolases (GH70) 3 (10). The GH70 family belongs to the clan GH-H, together with families GH13 and GH77. To date, 279 ␣-transglucosylase sequences have been reported in the CAZy database, and only 25% has been functionally characterized (11). Furthermore, only four three-dimensional structures of GH70 enzymes have been solved (12)(13)(14)(15). The enzymes are organized into five different structural domains, A, B, C, IV, and V. The A, B, and C domains are related to GH13 family enzymes, whereas domains IV and V are GH70specific domains. The fold of these enzymes is unique, and except for domain C, all domains are formed by non-contiguous fragments of sequences that assemble around a U-turn (15). These enzymes are ␣-retaining enzymes. They mainly catalyze ␣-transglucosylation reactions through a classical double displacement mechanism. The reaction starts with the formation of a ␤-D-glucosyl-enzyme intermediate involving a nucleophilic aspartate, an acid-base glutamic acid, and a third aspartate that stabilizes the covalent intermediate. In a second step the intermediate can be intercepted by a water molecule (hydrolysis reaction) or by the hydroxyl group of an acceptor (transgluco-sylation). The synthesis of the polymer occurs via successive transfers of glucosyl units onto the non-reducing extremity of the growing ␣-glucan chain (Scheme 1) (16 -18).
The formation of the regular hyper-branched dextran of L. citreum NRRL B-742 is even more intriguing. Several attempts to isolate the GH70 enzymes of L. citreum NRRL B-742 were reported. However, although the enzyme responsible for the production of dextran with both ␣-(136) and ␣-(134) linkages was isolated, the enzyme contributing to the comb-like structure was never identified (5). Of note, most of the GH70 enzyme activities were found strongly attached to the cell wall, making the enzymes difficult to purify (7). Crude enzyme extracts or sucrose fermentation broths were thus used to produce ␣-(133)-branched glucooligosaccharides from sucrose and maltose (7), and these compounds were shown to possess interesting functional properties. In particular, they stimulated in vitro growth of beneficial bacteria such as bifidobacteria or lactobacilli and inhibited the development of pathogens such as Salmonella sp. or Escherichia coli (23,24). In 2000, Kim and Robyt (25) cloned a gene from the mutant strain L. citreum NRRL B-742CB. The encoded protein, DSR-B-742, shared Ͼ95% amino acid identity with DSR-B dextransucrase from L. citreum NRRL B-1299 and synthesized a quasilinear dextran with Ͼ95% of ␣-(136) linkages (26). Thus far the mode of synthesis of the comb-like dextran has remained unknown.
A genomic approach was undertaken to gain insight into the formation of this hyper-branched ␣-glucan and to have access to enzymatic tools of interest for functional glucooligosaccharide synthesis. First, the genome of L. citreum NRRL B-742 was sequenced (27) and mined to determine its content in GH70 ␣-transglucosylase encoding genes. One putative GH70 gene SCHEME 1. Main reactions catalyzed by GH70 sucrose-active enzymes.

First ␣-(133) Branching Sucrases from the GH70 Family
showing divergence from various consensus sequences of GH70 family enzymes was cloned for recombinant expression in E. coli. The corresponding protein, named BRS-B, is an ␣-(133) branching sucrase. A shorter version of BRS-B, named BRS-B-⌬1, was rationally designed and characterized. The specificity of BRS-B and its truncated form was investigated and revealed the interest of these enzymes for the production of ␣-(133)-branched dextrans with a controlled degree of branching. In addition, comparison of these enzymes with the recently discovered ␣-(132) branching sucrase (BRS-A) allowed the identification of two enzymes in public databases that were found to catalyze dextran branching and were clustered with previously characterized branching sucrases in the GH70 phylogenetic tree.

Isolation of brsB Gene and Analysis of BRS-B Primary Structure
The brsB gene was identified in the L. citreum NRRL B-742 genome (27) (NCBI reference sequence: NZ_CCNG01000013.1) by performing a nucleotide BLAST against a homemade GH70 ␣-transglucosylase-encoding gene database. The protein sequence of BRS-B is deposited under the GenBank TM accession number CDX65123.1. Protein sequence comparison was performed either using protein Blast (NCBI) against the non-redundant protein sequence database (nr) for local alignments or using Clustal Omega for global alignments with other GH70 enzymes. Protein sequence global alignment with GTF-180-⌬N glucansucrase (15) and ⌬N 123 -GBD-CD2 (12) was used to determine length and position of domains A, B, C, IV, and V. The signal peptide cleavage was predicted using SignalP 4.1 server.

Isolation of Genomic DNA
L. citreum NRRL B-742 was grown on 1.5 ml of MRS medium (Sigma-Aldrich) at 30°C overnight. Genomic DNA was isolated using the Wizard genomic DNA purification kit (Promega, Madison, WI) according to the supplier's recommendations for Gram-positive bacteria.

Cloning of brsB and brsB⌬1 Genes
The gene brsB and a truncated form lacking the signal peptide and the last 575 amino acids at the C terminus of the protein (brsB⌬1) were amplified by PCR from L. citreum NRRL B-742 genomic DNA template. The primers used are described in Table 1. The addition of the CACC sequence (underlined) to the 5Ј-forward primers allowed the correct insertion of the genes into the pENTR/D-TOPO vector (Life Technologies). From a positive entry clone, LR recombination (Gateway LR Clonase II enzyme mix, Life Technologies) was performed using the destination vectors described in Table 1. Expression clones were selected on LB agar plates supplemented with 100 g ml Ϫ1 of ampicillin. Plasmids were extracted with the Sigma GenElute HP Plasmid Miniprep kit, verified by restriction analyses and sequencing (GATC Biotech, Constance, Germany).
Erlenmeyer Flask Cultures-One liter of modified ZYM5052 medium (28) contained (i) 0% lactose, 0% glucose, 0.5% glycerol and 0.01% L-arabinose for BRS-B production or (ii) 0.1% lactose, 0% glucose and 1% glycerol for BRS-B-⌬1 production. Both were supplemented with ampicillin (100 g ml Ϫ1 ) and were inoculated with the corresponding starter culture at an A 600 nm of 0.05. Cultures were incubated at 21°C under agitation (150 rpm). After a 26-h incubation, cells were harvested by centrifugation, dispersed in 50 mM sodium acetate buffer, pH 5.75, at a final A 600 nm 80 and disrupted by sonication. The recombinant enzymes were recovered in the soluble fraction after centrifugation (15,000 ϫ g, 30 min, 4°C) of the crude cell extract.

Affinity Chromatography Purification of Enzymes
All purification assays were performed using the Ä KTAxpress (GE Healthcare) at 12°C. For His 6 tag affinity chromatography, cells were centrifuged and resuspended in binding buffer (20 mM phosphate sodium buffer, pH 7.4, 500 mM NaCl, 20 mM imidazole, 2.5% (v/v) glycerol) at a final A 600 nm of 200. After disruption by sonication, centrifugation (30,000 ϫ g, 30 min, 4°C), and filtration through a 0.22-m cartridge (Sartorius Stedim Biotech, Aubagne, France), lysates were applied onto a 1-ml HisTrap HP column (GE Healthcare) that had been equilibrated with the binding buffer. The proteins were eluted by an imidazole gradient from 10 to 500 mM over 25 min. Eluate fractions of 3 ml were desalted onto a 10-DG column (Bio-Rad) with 50 mM sodium acetate buffer at pH 5.75 with 100 mM NaCl, 0.05 g liter Ϫ1 CaCl 2 , and 0.1% (v/v) Tween 80. In order to remove small contaminants, purified eluates were washed 3 times onto Vivaspin 2 with a molecular weight cut-off of 100 (Sartorius Stedim Biotech).

Enzymatic Activity Assay
One unit of BRS-B, BRS-B-⌬1, BRS-C, or BRS-D is defined as the amount of enzyme that catalyzes the production of 1 mol of fructose/min at 30°C in 50 mM sodium acetate buffer at pH 5.75 from 292 mM sucrose. The enzyme activities were determined by measuring the amount of reducing sugars using the dinitrosalicylic acid method (29).

pH and Temperature Optima; Effect of Divalent Ions
Optimal pH and temperature were determined in duplicate using 1 unit⅐ml Ϫ1 of enzyme in the presence of 292 mM sucrose and 119 mM dextran (68,400 g⅐mol Ϫ1 ) for both BRS-B (in sonicated extract) and the pure BRS-B-⌬1. To evaluate the optimal pH, the assays were performed at 30°C in 50 mM citrate phosphate buffer with pH values varying from 3.5 to 7.0 in increments of 0.5 units. Optimal temperature was determined from assays carried out in 50 mM sodium acetate buffer at pH 5.75 at temperatures of 23, 26, 30, 33, 37, 40, or 45°C. The effect of salts on transferase activity of BRS-B-⌬1 was determined by incubating the enzyme (1 unit⅐ml Ϫ1 ) with 1 mM EDTA to trap the possible presence of any divalent ion in the catalytic core of the enzyme or with 1 mM CaCl 2 , CuCl 2 , ZnCl 2 , or MgCl 2 salts. Activity assays were performed at 30°C in 50 mM sodium acetate buffer, pH 5.75, 292 mM sucrose, and 119 mM of dextran (68,400 g⅐mol Ϫ1 ). For all these experiments, the enzymatic activity was determined by measuring fructose release by high performance liquid chromatography (HPLC) on a Dionex system and using an Aminex HPX-87K carbohydrate analysis column (300 ϫ 7.8 mm; Bio-Rad). Column oven temperature was set at 65°C, and ultrapure water was used as eluent (0.6 ml min Ϫ1 ).

Determination of BRS-B-⌬1 Kinetic Parameters
Reaction with Sucrose-Kinetic parameters with sucrose as the sole substrate were determined in 50 mM sodium acetate buffer at pH 5.75 supplemented with 250 g ml Ϫ1 bovine serum albumin, 23 g ml Ϫ1 of purified BRS-B-⌬1 at 30°C and using initial sucrose concentrations ranging from 3 to 600 mM. All initial velocities were measured by quantifying fructose release at the early stage of reaction (when Ͻ5% of the substrate is consumed and when leucrose and other oligosaccharide production can be neglected). At regular time intervals samples were withdrawn and immediately heated at 95°C for 5 min to stop the reaction. Fructose concentration was determined by HPLC as described above. K m , K i , and V max parameters were determined using SigmaPlot Kinectics software and fitted to a non-competitive substrate inhibition model.
␣-(133) Transglucosylation-Kinetic parameters for ␣-(13 3) transglucosylation were determined at 30°C in 50 mM sodium acetate buffer, pH 5.75, supplemented with 250 g ml Ϫ1 BSA, 9.5 mg liter Ϫ1 of enzyme, 300 mM sucrose, and dextran of 68,400 g⅐mol Ϫ1 at final concentrations ranging from 31 to 617 mM. For the sake of clarity, dextran concentrations will be expressed as molar concentrations of anhydroglucosyl units within the dextran, as determined by dividing dextran mass concentrations by 162 g⅐mol Ϫ1 . All initial velocities were measured when Ͻ5% of the sucrose was consumed. Initial ␣-(133) transglucosylation velocities were calculated by subtracting fructose (representing the total activity) by free glucose (representing hydrolase activity) production rates. Glucose and fructose concentrations were determined by HPLC using the same conditions as those described previously. K mdextran and V max were estimated using SigmaPlot Kinetics software and fitted to a Michaelis-Menten model.

HPAEC-PAD Analysis
To determine production yield and assay sucrose depletion, reaction media were analyzed by HPAEC-PAD (high performance anion exchange chromatography with pulsed ampero-First ␣-(133) Branching Sucrases from the GH70 Family metric detection) using a CarboPac TM PA100 (4 ϫ 250 mm) analytical column coupled with a CarboPac TM PA-100 guard (4 ϫ 50 mm). Glucose, fructose, leucrose, and sucrose were separated using a 30-min sodium acetate gradient from 6 to 300 mM in 150 mM NaOH. The amount of each sugar was determined using standards of 5, 10, 15, and 20 mg kg Ϫ1 . Samples were prepared in ultra-pure water and diluted to a final concentration of 10 -15 mg kg Ϫ1 of total sugars for quantitative analyses and a final concentration of 250 mg⅐kg Ϫ1 (total sugars) for qualitative analyses. The percentage of glucosyl units from sucrose incorporated into glucose (%G glucose ) or leucrose (%G leucrose ) was calculated as follows.

High Performance Size Exclusion Chromatography (HPSEC) Analysis
HPSEC analyses were carried out using Shodex OH-Pak SB-802.5 and SB-805 columns in series coupled with a Shodex OH-Pak SB-G guard column and placed in an oven at 70°C. Elution was performed using 0.45 M NaNO 3 and 1% ethylene glycol (v/v) as the eluent using a flow rate of 0.3 ml min Ϫ1 . The samples were diluted at 10 g kg Ϫ1 of total sugars and also in 0.45 M NaNO 3 and 1% ethylene glycol.
The percentage of glucosyl units from sucrose incorporated into ␣-(133)-branched dextran was calculated as, The percentage of glucosyl units from sucrose incorporated into small oligosaccharides was deduced as, Here %G dextran , %G glucose , and %G leucrose have been determined from HPSEC and HPAEC-PAD analyses.

H and 13 C NMR Analysis of Oligosaccharides and Polymers
To determine ␣-(133) linkage content, freeze-dried polymer samples were dispersed in deuterated water at a final concentration of 20 mg ml Ϫ1 . NMR spectra were recorded on an Advance 500 MHz spectrometer (Bruker) operating at 500 MHz for 1 H NMR and 125 MHz for 13 C using a 5-mm z-gradient TBI probe. The data were processed using TopSpin 3 software. One-dimensional 1 H NMR spectra was acquired by using a zgpr pulse sequence (with water suppression). All measurements were performed at 298 K, and chemical shifts were referenced to an internal reference sodium 2,2,3,3-tetradeutero-3-trimethylsilylpropanoate ( 1 H ϭ 0 ppm) and acetone ( 13 C ϭ 31.08 ppm). The percentages of ␣(133) and ␣(136) linkages in ␣-glucans were calculated by integrating the corresponding anomeric proton signals (30).

Model Construction
Sequence alignment between BRS-B (from residues Ile-220 to Ile-1300) and ⌬N123-GBD-CD2 was generated using MUS-CLE (31). This alignment was checked, locally corrected in variable loop regions, and submitted to SWISS-MODEL online service for protein structure prediction using the alignment mode (32). The same procedure was applied for the sequence of BRS-A (from residues Gly-223 to Val-1289). BRS-B and ⌬N123-GBD-CD2 share 53% sequence identity; BRS-A and ⌬N123-GBD-CD2 share 60% sequence identity. All drawings were realized using PyMOL software (DeLano Scientific).

Identification of New Branching Sucrases
Blast analysis (blastp and blastn) were performed to identify new putative branching sucrases in NCBI public databases. The putative proteins displaying similarities with characterized branching sucrases (i.e. BRS-A, BRS-B, and/or GBD-CD2) were selected considering that the sequence should display (i) one phenylalanine at subsite ϩ1

Cloning of the brsC Gene
First, a synthetic brsC gene (contig GenBank TM accession number AEIZ01000002, positions 68,715-63,391 on reverse strand) was designed to optimize its expression in E. coli (Gene-Cust, Dudelange, Luxembourg). The gene was then amplified by PCR from pUC57/brsC plasmid DNA template using the primers described in Table 1. The PCR product was then inserted into the pENTR/D-TOPO vector (Life Technologies). From a positive entry clone, LR recombination (Gateway LR Clonase II enzyme mix; Life Technologies) was performed with the pET-55-DEST destination vector (Merck Millipore). Expression clones were selected on LB agar plates supplemented with 100 g ml Ϫ1 of ampicillin. Plasmids were then extracted using the GenElute HP Plasmid Miniprep kit (Sigma), verified by restriction analyses, and sequenced (GATC Biotech). E. coli TOP10 competent cells (Life Technologies) were used for all cloning experiments.

Recombinant Production of BRS-C and BRS-D Branching Sucrases
The construction pET28b/brD was provided by Biomatik (Cambridge, ON, Canada) where the brsD gene (GenBank TM accession number AZBY01000038.1, locus tag LAKU_38c00010) was designed to optimize the heterologous expression in E. coli.
Starter Culture-E. coli BL21 star DE3 cells were freshly transformed by pET-55/brsC, pET28b/brD. 30 ml of LB medium supplemented with ampicillin (100 g ml Ϫ1 ) were inoculated with 200 l of transformation mix and incubated overnight at 37°C under agitation (200 rpm).

JOURNAL OF BIOLOGICAL CHEMISTRY 7691
Erlenmeyer Flask Cultures-1 liter of modified ZYM5052 medium (28) (with the changes 0.1% lactose, 0% glucose, and 1% glycerol, supplemented with ampicillin (100 g ml Ϫ1 ) was inoculated with the starter culture at an A 600 nm of 0.05. Cultures were incubated at 21°C under agitation (150 rpm). After 26 h of incubation, cells were harvested by centrifugation, resuspended in 50 mM sodium acetate buffer, pH 5.75, at a final A 600 nm of 80, and disrupted by sonication. Recombinant enzymes were recovered after centrifugation (15,000 ϫ g, 30 min, 4°C) in the soluble fraction of the crude cell extract.

Phylogenetic Analysis
A phylogenetic tree was constructed from 60 protein sequences, including the functionally characterized GH70 enzymes indexed in the CAZY database and the branching sucrases BRS-A, BRS-B, BRS-C, BRS-D, and GBD-CD2 . Protein sequences were aligned using Clustal Omega. GBlock program was used to remove highly variable regions. A total of 1371 positions (42%) were then conserved. The tree was constructed using the Neighbor-joining method using the SeaView 4.5.4 software. The percentage of replicate trees in which the associated taxa clustered together in the bootstrap test (100 replicates) is shown next to the branches. The tree is drawn to scale, with branch lengths in the same units as those of the evolutionary distances used to infer the phylogenetic tree. The evolutionary distances were computed using the Poisson correction method and are in the units of the number of amino acid substitutions per site. The tree was displayed using FigTree v1.4.2 software.
For BRS-B, a signal peptide with a cleavage site between positions 39 and 40 was predicted using SignalP 4.1 server. Based on sequence alignment, the protein was predicted to belong to the GH70 enzyme family and to adopt the same structure as those displayed by other GH70 enzymes (12)(13)(14), which consist of five structural domains A, B, C, IV, and V, of which the limits were assigned by comparison with those of GTF180-⌬N and ⌬N 123 GBD-CD2 (12,15) (Fig. 2). The three catalytically important amino acids involved in the formation of the ␤-D-glucosylenzyme intermediate are conserved in the BRS-B sequence and are located in catalytic domain A. By analogy with other GH70 ␣-transglucosylases, Asp-671, Glu-709, and Asp-788 residues are predicted to play the role of the nucleophile, the acid/base catalyst, and the transition state stabilizer, respectively. Domain V is formed by two fragments, one at the N terminus and  First ␣-(133) Branching Sucrases from the GH70 Family the other at the C-terminal extremity of the protein, and is predicted to constitute both extremities of the U-shape fold (15). Like most of the domains V of GH70 enzymes, this domain is rich in YG repeats that were previously suggested to play a key role in interactions between glucansucrases and their polymers (36). Eight and two YG repeats that could be involved in ␣-glucan binding were identified at the N-and C-terminal extremities, respectively (Fig. 2). Additionally, at the C terminus, 6 APY repeats were identified and shared 99% of identity with those found at the C-terminal ends of the alternansucrase (ASR) from L. citreum NRRL B-1355 and the branching sucrase BRS-A from L. citreum NRRL B-1299 (513 out of 518 amino acids) (37,38). The APY repeats of alternansucrase (ASR) were previously deleted without causing any alteration of enzyme activity or specificity (39).
The brsB gene was cloned into pBAD-DEST49 vectors and expressed in E. coli BL21-AI host strain to characterize the corresponding protein. The level of production reached 700 units⅐liter Ϫ1 of culture. Western blot analyses confirmed that the full-length BRS-B was the most abundant protein, even if slightly degraded forms could also be observed (data not shown).
As the native E. coli BL21-AI does not produce any sucroseactive enzymes, the sonication extract was used to characterize the catalytic activities of BRS-B. The optimal temperature and pH were estimated at 30 -33°C and pH 5.5-6 (data not shown). All subsequent assays were thus performed at 30°C and at pH 5.75. Enzymatic reactions were first carried out in the presence of 292 mM sucrose, the natural substrate for most of GH70 enzymes. The main reaction products formed were glucose and leucrose (O-␣-D-glucopyranosyl-(135)-D-fructopyranose). A total of 51% of the glucosyl units from sucrose were incorporated into leucrose, and 33% were transferred onto water (hydrolysis reaction). The remaining part, representing 16% of the transferable glucosyl units, was incorporated into various small size glucooligosaccharides, which were not isolated and characterized due to their very low level of production (supplemental Fig. S1). No trace of high molar mass ␣-glucans was detected using HPSEC-RI, indicating that BRS-B is not an efficient polymerase.
Acceptor reactions were carried out using 1500 g⅐mol Ϫ1 dextran as the acceptor (66.6 mM) and sucrose as the donor (292 mM). As seen on HPAEC-PAD chromatograms, the product profile rapidly changed during the course of the reaction (Fig.  3a). The intensity of the peaks corresponding to linear isomaltooligosaccharides decreased, whereas new signals were clearly visible. The products synthesized after total sucrose depletion were analyzed by NMR spectroscopy. The 1 H NMR spectrum showed two peaks assigned to anomeric resonances at 4.99 ppm and 5.33 ppm, which correspond to the anomeric proton of a glucosyl unit engaged in ␣-(136) or ␣-(133) linkages, respectively. Integration of the anomeric signals showed that the amount of ␣-(133) linkages in dextran reached 39% that of the total osidic linkages (Fig. 3b). These results demonstrate that BRS-B catalyzes the transglucosylation reaction from sucrose onto dextran acceptors through the formation of ␣-(133) linkages.
Several dextrans of higher molar mass between 39,100 g⅐mol Ϫ1 and 2 ϫ 10 6 g⅐mol Ϫ1 were also tested as acceptors. All of them were efficiently branched by BRS-B, as evidenced by the percentage of ␣-(133) linkages reaching a maximum of 50%, indicating that each glucosyl moiety of the ␣-(136) backbone chain was substituted by an ␣-(133)-linked glucosyl residue (Fig. 4a).
The J-Mod spectrum of the dextran shows two anomeric 13 C chemical shifts at 98.63 ppm and 100.30 ppm corresponding to the 1 H anomeric signals at 4.98 ppm and 5.33 ppm of the ␣-(136) and ␣-(133)-linked glucosyl units, respectively. The presence of ␣-(133)-linked branching was further attested by the typical signal at 81.37 ppm, which was assigned to C-3 of the 3,6-O-disubstituted glucopyranosyl units (9,40). Only one type of resonance is observed at 100.30 ppm, indicating that the ␣-(133) linkages of the dextran are solely branching linkages. This is confirmed by the presence of only one chemical shift at 61.20 ppm attributed to free C6. At 66.33 ppm, the resonances are attributed to bound C6 of the 3,6-O-disubstituted and 6-Odisubstituted glucosyl units of the main chain (Fig. 4b).
To date, BRS-B is the first reported ␣-(133) branching sucrase. Its ability to branch high molar mass dextran up to 50% suggests that BRS-B is likely involved in the synthesis of L. citreum NRRL B-742 comb-like dextran. Based on sequence alignment, the other putative GH70 enzymes encoded by L. citreum NRRL B-742 are predicted to be dextransucrases synthesizing linear dextrans. Dextransucrase activities were previously isolated in the fermentation broth of L. citreum NRRL B-742 (5,22,26). It can thus be assumed that the synthesis of the hyper-branched dextran is due to the synergistic action of at least one dextransucrase and the branching enzyme, BRS-B. The dextransucrase would synthesize the ␣-(136) backbone, and the ␣-(133) branching sucrase would be involved in branching via ␣-(133) osidic bond formation. There are several questions that remain, including understanding which dextransucrases of L. citreum NRRL B-742 work in tandem with BSR-B. Additionally, it is not known whether or not elongation and branching are concomitant. These pending questions require additional experiments, such as RT-quantitative PCR studies, secretome analysis at regular intervals of the culture, and knock-out of either polymerases or branching sucrase encoding genes combined with structural analyses of the produced dextran.

Construction of BRS-B-⌬1, a Truncated Form of BRS-B-
Despite several attempts, the purification of BRS-B did not give satisfying results due to very weak protein binding onto nickel and cobalt affinity resins. This was attributed to an inadequate exposure of the tags, which could prevent correct interaction with the resin. To facilitate the purification process, a shorter form deleted of the signal peptide and all APY motifs at the C-terminal end, was constructed (Fig. 2). Notably, 8 of the 10 identified YG repeats were conserved. The position of the truncations at the C-and N terminus ends were determined from sequence alignment with the ␣-(132) branching sucrase ⌬N 123 GBD-CD2. The production of the truncated protein BRS-B-⌬1 was optimized to a level of 1200 units⅐liter Ϫ1 using pET-55-DEST and BL21 Star DE3 E. coli strain. Western blot analysis showed that the protein was not degraded at the N-terminal First ␣-(133) Branching Sucrases from the GH70 Family extremity. The C-terminal His 6 tag was then used to purify the enzyme onto Ni-NTA affinity resin. The purification yield reached 85%, with a corresponding purification factor of 17. BRS-B-⌬1-specific activity on sucrose (292 mM) was estimated at 24 units⅐mg Ϫ1 of protein. The optimal temperature and pH were determined to be between 33 and 40°C and 5-5.75, respectively (Fig. 5, a and b) and are in the same range as those observed for the non-purified enzyme BRS-B. A slight improvement of transferase activity (near 6%) was observed with the addition of 1 mM Ca 2ϩ (Fig. 5c), as previously described for the branching sucrase ⌬N123-GBD-CD2 (increase of 13% (12)). In contrast, Zn 2ϩ ions had the most deleterious effect on BRS-B-⌬1 activity, which decreased by 23% in the presence of Zn 2ϩ ions (Fig. 5c). From a sucrose substrate, BRS-B-⌬1 showed a slightly higher hydrolytic activity than BRS-B and produced 42% glucose, 48% leucrose, and 10% of small oligosaccharides. The profiles of products obtained with dextran acceptors were similar to those obtained with the full-length protein (data not shown). Of note, the deletion of the APY repeats at the C terminus extremity of the protein did not affect the enzyme specificity and efficiency, as previously observed with the alternansucrase and inulosucrase (39,41).   The kinetic parameters of the recombinant protein BRS-B-⌬1 were determined with and without dextran acceptor. Without the acceptor, sucrose hydrolysis was inhibited by Ͼ150 mM sucrose, and the kinetic data fitted a non-competitive substrate inhibition modeled by Equation 1 with a K m , K i , and k cat values estimated at 7.0 Ϯ 0.7 mM, 595 Ϯ 88 mM, and at 81.4 s Ϫ1 , respectively. An initial rate of transglucosylation was determined at a fixed sucrose concentration (300 mM) and using dextran acceptor (68,400 g⅐mol Ϫ1 ) concentrations varying from 31 to 617 mM (in anhydroglucosyl unit equivalents). A classical Michaelis-Menten profile was obtained with K m dextran and k cat values of 0.27 Ϯ 0.08 mM and 949 s Ϫ1 , respectively. A 12-fold increase of the catalytic constant was observed compared with the reaction without dextran ( Table 2). The formation of ␣-(133) branches onto dextran molecules is very efficient (k cat /K mdextran ϭ 3514 s Ϫ1 mM Ϫ1 ), and the k cat value is close to that reported for ␣-(132) branching formation catalyzed by the ␣-(132) branching sucrase GBD-CD2 (20) ( Table  2). Such values range among the highest k cat values reported for GH70 members.
Synthesis of Dextrans with Controlled Amounts of ␣-(133) Linkages-The opportunity to synthesize different ␣-glucan structures with a controlled degree of branching using BRS-B-⌬1 was investigated in more details. To this end, low and high molar mass dextrans (1500 and 2 ϫ 10 6 g⅐mol Ϫ1 , respectively) were tested as acceptors at different sucrose:acceptor ratios. When sucrose was in large excess, the production yields of ␣-(133)-branched dextrans decreased in favor of the formation of side products (mainly glucose and leucrose, production of small oligosaccharides being negligible). In contrast, when the dextran concentration increased, glucose and leucrose production was reduced (Fig. 6a). The percentage of ␣-(133) link-ages in branched dextrans was tightly controlled and evolved from 5 to 50% by adjusting the sucrose:acceptor ratio (Fig. 6b). This parameter was also shown to be critical for controlling the percentage of ␣-(132) branching catalyzed by the ␣-(132) branching sucrase, GBD-CD2 (20). An initial sucrose/dextran ratio higher than three has to be used to obtain a highly branched dextran. By controlling the sucrose:acceptor ratio and the size of the dextran acceptor, a large array of products can be produced, opening the possibility of new applications for these kinds of molecules.
Comparison of BRS-B with Other Branching Sucrases and Glucansucrases-The four functional sequence signatures (motifs I-IV), usually well conserved in the GH70 family and also found in related the GH13 family (10), were aligned with the corresponding motifs found in the characterized GH70 enzymes from other organisms (Table 3). Interestingly, BRS-B exhibits numerous sequence similarities with GBD-CD2 and BRS-A from L. citreum NRRL B-1299, two branching sucrases (19,21). In particular, Phe-675 of motif II and Ile-783, His-785, Lys-789, and Val-795 of motif IV are only conserved in the branching sucrases. They are not found in other GH70 conserved motifs and could thus be suggested to be important for branching activity. Other amino acids are uniquely found in BSR-B. They include residues 673-IS of motif II, 711-PKGE of motif 796-IH of motif IV, and F1184 of motif I ( Table 3).
The three-dimensional model of BRS-B was constructed by homology modeling using ⌬N 123 -GBD-CD2 branching sucrase as template (Fig. 7). The model was compared with the 3D-structures of ⌬N 123 -GBD-CD2 (PDB: 3TTQ) and with the inactive mutant of GTF-180-⌬N in complex with sucrose substrate (PDB: 3HZ3). According to the model, BRS-B also adopts a U-shape fold organized in five domains. As shown in Fig. 7,      -675 (Fig. 8a). Notably, one phenylalanine was also found in ⌬N 123 -GBD-CD2, and was shown to be critical for dextran branching (12). Moreover, comparison of BRS-B and GTF-180 backbone, also revealed a rearrangement around residues A710 and P711that was observed in ⌬N 123 -GBD-CD2. This rearrangement was previously suggested to be responsible for a weaker binding of the fructosyl ring in branching sucrases (12). The identification of the determinants for ␣-(133) branching specificity remains difficult in the absence of three-dimensional structures of BRS-B in apo form or in complex with sucrose or  First ␣-(133) Branching Sucrases from the GH70 Family dextran. However, according to the structural comparison of branching enzymes and polymerases, amino acids can be pointed out that constitute interesting targets for mutation to assess their role in the ␣-(133) branching specificity.
Searching for New Branching Sucrases in Public Databases-The emergence of natural branching activities through the isolation and characterization of BRS-A (21) and BRS-B in L. citreum raised the question of the representation of such type of activity in lactic acid bacteria clade. To shed some light on this issue, a blast search against public sequence databases were carried out to ascertain which sequences contain the residues uniquely conserved in branching enzyme catalytic cores. Two new putative branching sucrases (renamed BRS-C and BRS-D) were identified, one in Leuconostoc fallax KCTC3537 and the other in Lactobacillus kunkeei EFB6. The percentage of amino acid identity and similarity with BRS-A and BRS-B is detailed in supporting information (supplemental Table S2), and the alignment of the conserved motifs is presented in Table 3. Based on the sequence alignment, the two putative branching sucrases are predicted to possess the catalytic triad DED and the five structural domains, A, B, C, IV, and V, specific to the GH70 family enzymes (supplemental Fig. S2). Similar to the characterized BRS-A (21) and BRS-B, YG repeats (36) were mainly identified at the N-terminal extremity of the protein. All newly identified genes encoding putative branching sucrases are located on the reverse strand of their genome, as previously observed for brsA in L. citreum NRRL B-1299 genome (21) and brsB in L. citreum NRRL B-742 genome. All of them are organized in tandem with another putative glucansucrase (supplemental Fig. S3). First ␣-(133) Branching Sucrases from the GH70 Family APRIL 1, 2016 • VOLUME 291 • NUMBER 14

JOURNAL OF BIOLOGICAL CHEMISTRY 7699
Characterization of Two New Putative Branching Sucrases and Phylogenetic Analysis-The genes encoding the putative branching sucrases were expressed in E. coli. The production levels of the two new proteins are reported in supplemental Table S3. As already observed for BRS-A, BRS-B, and GBD-CD2, the putative BRS-C and BRS-D mainly produced glucose, leucrose, and a low proportion of small oligosaccharides from sucrose substrate (for detailed production yields, see supplemental Table S4). Acceptor reactions were then performed using sucrose and a small dextran of 1500 g⅐mol Ϫ1 as an acceptor. At the end of the reaction, HPAEC-PAD and H 1 NMR analyses demonstrated that dextran was modified by BRS-C and BRS-D (Fig. 9). BRS-C from L. fallax KCTC 3537 catalyzed ␣-(133) transglucosylation from sucrose to dextrans, whereas BRS-D from L. kunkeei EFB6 was specific for ␣-(132) transglucosylation. These findings reveal for the first time that the branching specificity is not restricted to L. citreum sp. but is also present in L. fallax as well as in L. kunkeei. Interestingly, the phylogenetic analyses show that the branching sucrases cluster together in a distinct branch (Fig. 10). In addition, the branching sucrase subtree is divided into two subgroups, one corresponding to enzymes showing an ␣-(132) linkage specificity and the other to enzymes with ␣-(133) linkage specificity.
The discovery of the new branching sucrases expands the natural repertoire of the characterized GH70 ␣-transglucosylases. For years the GH70 family has been mainly known to house glucansucrases. Recently, a new subgroup was disclosed that contains 4,6-␣ glucanotransferases, a type of ␣-transglucosylases that uses ␣-(134)-linked glucans as a donor instead of sucrose (42,43). According to our biochemical and phylogenetic analyses, branching sucrases could also constitute a new GH70 subgroup. Five enzymes of this group have now been characterized, including the engineered ␣-(132) branching sucrase GBD-CD2 (19,20), BRS-A (21), BRS-B, BRS-C, and BRS-D.
The route is now opened to promising structure-function relationship studies that will aim at identifying the structural determinants responsible for polymerization or branching ability and linkage specificity. Despite common structural motifs with GH70 glucansucrases, the branching sucrases also display clear distinctive features. According to sequence analyses, dif- First ␣-(133) Branching Sucrases from the GH70 Family ferences have been highlighted in the substrate binding subsite ϩ1, which can be used to rapidly sort out branching sucrases from genomic data. Mutational and structural studies will now be necessary to better define the signature of the branching specificity and to provide further insight into the branching mechanism. Finally, a broad variety of branched dextrans varying in terms of size, osidic linkage and branching degree can be easily synthesized using these enzymes, emphasizing the potential of GH70 enzymes to access new structures of biotechnological interest.