Molecular Cloning and Characterization of a β-l-Arabinobiosidase in Bifidobacterium longum That Belongs to a Novel Glycoside Hydrolase Family*

Extensin is a glycoprotein that is rich in hydroxyprolines linked to β-l-arabinofuranosides. In this study, we cloned a hypBA2 gene that encodes a novel β-l-arabinobiosidase from Bifidobacterium longum JCM 1217. This enzyme does not have any sequence similarity with other glycoside hydrolase families but has 38–98% identity to hypothetical proteins in Bifidobacterium and Xanthomonas strains. The recombinant enzyme liberated l-arabinofuranose (Araf)-β1,2-Araf disaccharide from carrot extensin, potato lectin, and Araf-β1,2-Araf-β1,2-Araf-β-Hyp (Ara3-Hyp) but not Araf-α1,3-Araf-β1,2-Araf-β1,2-Araf-β-Hyp (Ara4-Hyp) or Araf-β1,2-Araf-β-Hyp (Ara2-Hyp), which indicated that it was specific for unmodified Ara3-Hyp substrate. The enzyme also transglycosylated 1-alkanols with retention of the anomeric configuration. This is the first report of an enzyme that hydrolyzes Hyp-linked β-l-arabinofuranosides, which defines a new family of glycoside hydrolases, glycoside hydrolase family 121.

gum JCM 1217 on the basis of the sequence of BL0464, a hypothetical protein from B. longum NCC2705 (7), which contains C-terminal FIVAR domains that are involved in association with the bacterial cell wall (Pfam 07554). Similar to BL0464, BL0421, another hypothetical protein in B. longum, has three FIVAR domains (KEGG). BL0420, the gene that flanks BL0421, encodes a protein that also contains Cterminal FIVAR domains and belongs to glycoside hydrolase (GH) family 43, which includes xylosidases and arabinosidases. Schell et al. (8) predicted that BL0420 and BL0421 may be involved in binding or degrading xylan or hemicellulose. We hypothesized that these proteins are localized on the cell surface and coordinately degrade arabinan or xylan. Because BL0421 is not homologous to any known glycoside hydrolases, we predicted that the hypothetical protein is a novel enzyme.
Glycosidases that recognize the sugar chains of glycoproteins are useful in elucidating the presence and function of glycans. For instance, endo-␤-N-acetylglucosaminidase selectively cleaves N-linked glycans from glycoproteins. In this study, we cloned the gene of a BL0421 homolog from B. longum JCM 1217 and characterized the properties of the recombinant enzyme, which is a novel ␤-L-arabinobiosidase. This is the first report of an enzyme that catalyzes the degradation of Hyp-linked ␤-L-arabinofuranosides.

EXPERIMENTAL PROCEDURES
Substrates-p-Nitrophenyl (pNP) substrates were obtained from Sigma. Sugar beet arabinan, debranched arabinan, arabinoxylan, and rhamnogalacturonan were purchased from Megazyme and purified by ethanol precipitation. Potato lectin was purified from potato tubers as described previously (9), by using affinity chromatography with a Chitopearl Basic BL-01 (Fuji Spinning) chitin column. Extensin was prepared from carrot root as described previously (10).
The extensin was used to prepare Hyp-linked ␤-L-arabinofuranosides as follows. First, a sample of extensin was hydrolyzed for 16 h at 105°C with 0.22 M Ba(OH) 2 , and then the hydrolysate was neutralized with sulfuric acid and centrifuged at 15,000 ϫ g for 20 min. The supernatant was applied to a Dowex 50 W ϫ 4 resin (H ϩ form). After washing the resin with 10 mM acetate buffer (pH 2.5), the Hyp-linked ␤-L-arabinofuranosides were eluted with 1.5 M aqueous ammonia and then evaporated to dryness. Next, the dried residue was dissolved in water and applied onto a BioGel P2 column with distilled water. Finally, the eluate fractions containing Araf-␤1,2-Araf-␤-Hyp (Ara 2 -Hyp), Ara 3 -Hyp, and Ara 4 -Hyp were analyzed by matrix-assisted laser desorption/ionization timeof-flight mass spectrometry (MALDI-TOF MS).
Hyp and Ara 3 -Hyp were dansylated as described by Gray (11). The artificial cis form of Hyp-linked O-glycans was isomerized from the natural trans form by using an alkaline hydrolysis procedure (12). The cis and trans conformers were separated by high performance liquid chromatography (HPLC) on a Cosmosil 5C18-AR-II (10 ϫ 250 mm, Nacalai) column with a mobile phase of methanol and 10 mM sodium phosphate (pH 2.5) (60/40, v/v) and a constant flow rate (1.0 ml ⅐ min Ϫ1 ) at 30°C. The elution was monitored by a fluorescence detector (FP-202; JASCO) with excitation and emission wavelengths of 365 and 530 nm, respectively. Finally, the eluate was desalted using a Sep-Pak Plus tC18 cartridge.
Expression and Purification of Recombinant HypBA2-The genomic DNA of B. longum JCM 1217 was extracted by using a FastPure DNA kit (Takara) and then used to perform PCR amplification of the gene for the BL0421 homolog, HypBA2. The forward (5Ј-AGGAGATATACCATGGCCGATACGC-CGTCA-3Ј) and reverse (5Ј-TGCTCGAGTGCGGCCGCCG-GAAGATGAACC-3Ј) primers were designed from nucleotides 96 -112 and 4345-4357, respectively, of BL0421 from B. longum NCC2705. The underlined nucleotides represent the NcoI and NotI sites, respectively. The PCR amplification of hypBA2-C⌬486, which encodes amino acids 33-1464, was designed to eliminate the N-terminal signal peptide and Cterminal cell surface binding domain (residues 1465-1943). Then, the amplicon was digested with NcoI and NotI and inserted into the corresponding sites of the pET-23d vector (Novagen) to generate a C-terminal His 6 -tagged recombinant protein. The resulting pET23d-hypBA2-C⌬486 plasmid was transformed into Escherichia coli BL21 (DE3) cells and then grown at 20°C by using the Overnight Express Autoinduction System (Novagen). Subsequently, the cell cultures were centrifuged, and then the pellet was resuspended in BugBuster protein extraction reagent (Novagen). The His-tagged proteins were purified on a TALON metal affinity resin (Clontech), desalted by dialysis with a cellulose membrane (Wako), and concentrated by using a 10-kDa ultrafiltration membrane (Millipore). In addition, the PCR product of the full-length hypBA2 gene was sequenced on an ABI 3100 DNA sequencer using a Big-Dye Terminator 3.1 Cycle Sequencing kit (Applied Biosystems).
Analysis of HypBA2 Hydrolysis and Transglycosylation Products-Arabinan and debranched arabinan (1.6 mg) were incubated at 30°C for 2 h or overnight with 1.0 g of purified HypBA2 enzyme in 100 l of 50 mM sodium acetate buffer (pH 6.0) with or without alcohol. The reaction was stopped by boiling for 3 min and then precipitated with 80% ethanol. The supernatant was dried in a centrifugal concentrator and dissolved in 10 l of water and then analyzed by thin layer chromatography (TLC) on a Silica Gel 60 aluminum plate (Merck) using a 2:1:1 solvent mixture (v/v/v) of ethyl acetate/acetic acid/water. The sugars were visualized by spraying an orcinolsulfate reagent onto the plate (13).
For structural analysis of the transglycosylation product, 8 g of sugar beet arabinan was mixed with 100 g of HypBA2 enzyme in 500 ml of 50 mM sodium acetate buffer (pH 6.0) with 30% methanol and then incubated at 30°C overnight. Subsequently, the completed reaction was evaporated, redissolved in 200 ml of water, and precipitated with 80% ethanol. The supernatant was collected and evaporated to dryness. Afterward, the dried sample was dissolved in 1 ml of water and separated by HPLC on a Cosmosil Sugar-D (4.6 ϫ 250 mm, Nacalai) column at 30°C with a mobile phase of acetonitrile and water (75/25, v/v) and a constant flow rate (1.0 ml ⅐ min Ϫ1 ). The elution was monitored by a refractive index detector (RI-8022; TOSOH). Finally, the fraction that contained the transglycosylation product (ϳ1 mg) was collected and used for NMR analysis.
Assay of ␤-L-Arabinobiosidase Activity-The hydrolytic activity of the HypBA2 enzyme was assayed by using dansylated cis-Ara 3 -Hyp (Ara 3 -Hyp-DNS) as a substrate. The 40-l reaction mixture contained 100 mM sodium acetate buffer (pH 5.5), 35 M substrate, and 0.44 milliunits ⅐ ml Ϫ1 of the HypBA2 enzyme. One unit of enzyme activity was defined as the amount of enzyme that is needed to produce 1 mol of cis-Ara-Hyp-DNS ⅐ min Ϫ1 . The amount of cis-Ara-Hyp-DNS was quantified by measuring the arabinose content by trifluoroacetic acid hydrolysis and high performance anion-exchange chromatography with pulsed amperometric detection (HPAEC-PAD) analysis as described below. After incubating the reaction at 30°C, the reaction was stopped by adding 10 l of 10% trichloroacetic acid and then analyzed by HPLC. The sample was applied to a Cosmosil 5C18-AR-II (2.5 ϫ 250-mm, Nacalai) column at 30°C with a mobile phase of methanol and 20 mM sodium phosphate (pH 2.5) (60:40, v/v) and a constant flow rate (1.0 ml ⅐ min Ϫ1 ). The elution was detected by fluorescence.
Transglycosylation of Ara 3 -Hyp-The transglycosylation reactions were performed by incubating 10 nmol of Ara 3 -Hyp with 3.5 milliunits ⅐ ml Ϫ1 of the HypBA2 enzyme at 30°C overnight in 100 l of 50 mM sodium acetate buffer (pH 5.5) with 20% methanol, ethanol, or n-propyl alcohol as an acceptor. Subsequently, the reaction products were analyzed by TLC with a 2:1:1 solvent mixture of ethyl acetate/acetic acid/ water (v/v/v). The sugars were visualized by spraying an orcinol-sulfate reagent onto the plate (13).
Assays for pH, Temperature, and Ca 2ϩ Dependence-The pH dependence of enzyme activity was determined between pH 2.0 and 9.0 by using the following buffers: 50 mM sodium acetate (pH 3.5-6.0), 50 mM MES (pH 5.5-7.0), 50 mM sodium phosphate (pH 6.5-8.0), and 50 mM Tris-HCl (pH 7.5-9.0). The effect of temperature on enzyme activity was examined by using 50 mM sodium acetate buffer (pH 5.5) at 10 -60°C. The effect of Ca 2ϩ on the enzyme activity was determined with 1 mM CaCl 2 or EDTA.
Sugar Composition Analysis-The enzyme reaction product was incubated with 2 M trifluoroacetic acid for 1 h at 120°C. Then the hydrolyzed sugar chains were analyzed by HPAEC-PAD with a CarboPac PA10 column (Dionex) at a flow rate of 0.7 ml ⅐ min Ϫ1 , as described previously (14).
MS and NMR Analysis-electrospray ionization-TOF MS analysis was performed on a Mariner Biospectrometry Work station (Applied Biosystems). MALDI-TOF MS was performed on a Voyager-DE mass spectrometer (Applied Biosystems) in the positive ion mode with 2,5-dihydroxybenzoic acid as the matrix. 1 H and 13 C NMR spectra were measured with a JMM-ECA600KS spectrometer (JEOL). Two-dimensional correlation spectroscopy, heteronuclear multiple bond correlation, and heteronuclear multiple quantum coherence spectra also were used to assign chemical shifts and anomeric configurations of enzyme products.
Deletion Mutagenesis-The KOD-Plus Mutagenesis kit (Toyobo) was used to create nine deletion mutants of HypBA2 by using the primers shown in supplemental Table  S1. The deletion mutants were as follows: C⌬629 (amino acids The bacteria were grown in Gifu anaerobic medium broth (Nissui) for 4 days at 37°C under anaerobic conditions. The cell cultures were centrifuged at 15,000 ϫ g for 20 min, and then the pellets were washed with 50 mM sodium acetate buffer (pH 6.0). Subsequently, they were resuspended in 100 mM sodium acetate buffer (pH 6.0) and incubated with 0.25 mM Ara 3 -Hyp-DNS for 8 h at 37°C. Finally, the reactions were analyzed by TLC. The spots on the plates were developed with a 3:1:1 mixture (v/v/v) of n-butyl alcohol/acetic acid/water and then visualized with ultraviolet (UV) light.

Molecular Cloning, Expression, and Purification of the
HypBA2 Protein-HypBA2 consisted of 1,943 amino acid residues exhibiting 97.6% identity with that of BL0421. The recombinant HypBA2-C⌬486 protein was expressed at 20°C as a soluble protein. SDS-PAGE showed that the purified recombinant HypBA2 protein migrated as a single band with an apparent molecular mass of 157 kDa (supplemental Fig. S1), which was in agreement with its calculated molecular mass of 156,797 Da.
Identification of the Liberated Carbohydrate from Arabinan-To identify the substrate for HypBA2, we screened its hydrolysis of various synthetic pNP substrates and natural hemicelluloses. The enzyme did not have any activity with pNP-␣-L-arabinofuranoside, pNP-␣-L-arabinopyranoside, pNP-␤-L-arabinopyranoside, arabinoxylane, or rhamnogalacturonan (data not shown). However, it released a carbohydrate from arabinan and debranched arabinan (supplemental Fig. S2A), which was composed of arabinose (supplemental Fig. S2B). Electrospray ionization MS analysis revealed two molecular ion peaks at m/z 305.12 and 321.09, which corresponded to the sodium and potassium adducts of arabinobiose (calculated m/z 305.08 and 321.06 for C 10 H 18 O 9 Na [MϩNa] ϩ and C 10 H 18 O 9 K [MϩK] ϩ , respectively). Furthermore, transglycosylation products were detected in the presence of 20% methanol, ethanol, and n-propyl alcohol (supplemental Fig. S2C). HPLC analysis of the transglycosylation product in the presence of methanol is shown in supplemental Fig. S2D. Electrospray ionization MS analysis revealed a molecular ion peak at m/z 319.11, which is consistent with the sodium adduct of a transglycosylation product, namely, Ara 2 -OMe (calculated m/z 319.10 for C 11 H 20 O 9 Na). The structure of this product was determined by 1 H and 13 C NMR ( Fig. 1 and supplemental Table S2). The observed multiplicity of the anomeric proton (doublet at 4.86 ppm) and its coupling constant (J 1,2 ϭ 4.4 Hz) were consistent with the product being a methyl ␤-L-arabinofuranoside (15). In addition, two-dimen- Cloning and Characterization of a Novel ␤-L-Arabinobiosidase FEBRUARY 18, 2011 • VOLUME 286 • NUMBER 7 sional correlation spectroscopy, heteronuclear multiple bond correlation, and heteronuclear multiple quantum coherence spectra of the transglycosylation product specifically identified the product as Araf-␤1,2-Araf␤-OMe.
Although a terminal ␤-Araf residue was detected in olive arabinan (16), the Araf-␤1,2-Araf␤ residue has never been found in arabinan. In this study, only small amounts of the liberated saccharides (about 0.02%) were produced from commercially available sugar beet arabinan. Because most of the arabinofuranosyl residues in arabinan are in the ␣-anomeric configuration, we assumed that the Araf-␤1,2-Araf␤ residue was a contaminant from plant tissues.
Transglycosylation Activity of HypBA2-Many retaining glycoside hydrolases transglycosylate 1-alkanols, such as methanol, and produce methyl glycosides that are stable enough for NMR analysis (17). HypBA2 transglycosylated with Ara 3 -Hyp in the presence of 20% methanol and ethanol, but 20% n-propyl alcohol inhibited enzymatic activity (Fig. 4). The coupling constant (J 1,2 ϭ 4.4 Hz) in the 1 H NMR spectrum of the transglycosylation product from the reaction in 30% methanol was consistent with Araf-␤1,2-Araf-␤-OMe and a mechanism that retains the anomeric configuration.
Effects of cis-trans Substrate Conformations, pH, Temperature, and Ca 2ϩ on HypBA2 Activity-HypBA2 liberated Ara-Hyp-DNS from both cis-and trans-Ara 3 -Hyp-DNS (Fig. 5). The retention time (R t ) of trans-Ara-Hyp-DNS (R t ϭ 4.4 min) was longer than that of trans-Ara 3 -Hyp-DNS (R t ϭ 4.0 min) (Fig. 5A). Similarly, the retention time of cis-Ara-Hyp-DNS (R t ϭ 5.8 min) was longer than that of cis-Ara 3 -Hyp-DNS (R t ϭ 4.7 min) (Fig. 5B). Because the difference in retention times was larger for the cis form than that of the trans form, we used the cis form of Ara 3 -Hyp-DNS to characterize the enzyme.
The optimal temperature and pH for HypBA2 activity were 30°C and 5.5-6.0, respectively (supplemental Fig. S3). The K m and k cat values were calculated as 10.7 M and 2.7 s Ϫ1 , respectively. Furthermore, 1 mM Ca 2ϩ increased the enzy-matic activity by 43.8%, but 1 mM EDTA decreased the activity to 1.4% (Table 1).
Sequence Analysis of HypBA2-HypBA2 consisted of 1,943 amino acids that included a putative signal peptide, 3 F5/8 type C, 2 Big4, and 3 FIVAR domains (Fig. 6A). The F5/8 type C domain (Pfam 00754) is referred to as carbohydrate-binding module (CBM) family 32 that is generally involved in galactose binding. The Big4 domain is a bacterial Ig-like domain found in bacterial surface proteins (Pfam 07532). The amino acid sequence of HypBA2 was not similar to any known GH families. However, it was 38 -98% identical to other hypothetical proteins from Bifidobacterium, Xanthomonas, and actinomycetes ( Fig. 6B and supplemental Fig. S4). Interestingly, almost all hypBA2 homologs flank GH43 homologous genes (Fig. 6B).
Detection  Fig. S5). In addition, the arabinobiose product from     (Fig. 6). These results suggested that ␤-L-arabinobiosidase exists in B. longum and some other bifidobacterial species.

DISCUSSION
Glycoside hydrolases involved in the degradation of ␣-Larabinofurano-oligosaccharides were found in six GH families (GH3, GH43, GH51, GH54, GH62, and GH93) in the carbohydrate-active enzyme (CAZy) data base (18,19). In addition, glycoside hydrolases in the GH42 and GH27 families possess ␣and ␤-L-arabinopyranosidase activities, respectively (20,21). In this study, we cloned and characterized a hypothetical protein, HypBA2, and identified it as a novel ␤-Larabinobiosidase. This is the first report of an enzyme that hydrolyzes ␤-L-arabinofuranosides. HypBA2 is a retaining enzyme that has strict substrate specificity for Ara 3 -Hyp. Deletion mutants of HypBA2 revealed that the N-terminal conserved region (aa 33-907) and the adjacent F5/8 type C domain (aa 908 -1044) are critical for enzymatic activity. The homologous genes in Bifidobacterium, Xanthomonas, and actinomycetes also conserved this N-terminal region; however, it did not contain any known sequence motifs. Therefore, we propose that the enzyme be assigned to a new GH family 121. Bifidobacteria are strictly anaerobic lactic acid-producing bacteria that are common in the lower intestinal microflora of human and animals. Because simple digestive sugars are preferentially absorbed in the upper intestinal tract, these bacteria survive in the lower intestinal tract by using glycoside hydrolases to cleave complex oligosaccharides (22). For example, B. longum NCC2705 encodes 14 members of the GH43 and GH51 families including ␣-L-arabinofuranosidases (23), and monomeric carbohydrates, such as L-arabinose, induce ␣-Larabinofuranosidase activity in B. longum NIZO B667 (24). Several reports indicate the ability of B. longum to grow on L-arabinose and ␣-L-arabinofuranosides (25)(26)(27)(28). Furthermore, enzymes involved in L-arabinose metabolism to D-xylulose-5-phosphate are conserved in B. longum NCC2705: Larabinose isomerase (BL0272), L-ribulose-5-phosphate 4-epimerase (BL0273), and putative L-ribulokinase (BL0274). As a result, arabinosidases play important roles in the metabolism of B. longum by producing arabinose that can be used as carbon and energy sources.
The cell walls of dicot plants contain 33-75% of Ara 4 -Hyp in the total Hyp residues (6), and potato lectin contains 47% of Ara 4 -Hyp in the total Hyp residues (31). Ashford et al. (5) reported that fungus ␣-L-arabinofuranosidase hydrolyzed Ara 4 -Hyp to Ara 3 -Hyp. Because HypBA2 cannot hydrolyze Ara 4 -Hyp, ␣-L-arabinofuranosidase will be needed for effective degradation of ␤-L-arabinofuranosides. The hypBA2 gene forms a gene cluster with a GH43 homologous gene, which is conserved in many HypBA2 homologs (Fig. 6B). Therefore, HypBA2 and the neighboring GH43 protein may coordinately act on ␤-L-arabinofuranosides. Interestingly, BL0420 indicates highly homologous with other neighboring GH43 proteins (35% identity with that of Xanthomonas albineans). The gene cluster may be beneficial for the degradation of plant extensin, because many of Xanthomonas strains cause plant diseases. Bifidobacteria and Xanthomonas are phylogenetically distant enough. In addition, the gene cluster cannot found in other gut bacteria. Hehemann et al. (32) reported that the human gut bacteria acquired genes for porphyran degradation from a marine bacterium. Therefore, a part of Bifidobacteria probably received the gene cluster by horizontal gene transfer from Xanthomonas or actinomycetes.
Many lectins and CBMs require calcium to bind carbohydrates (33). CBM32 members (referred to as F5/8 type C) also have a binding calcium (34,35). All of the HypBA2 deletion mutants showed calcium-dependent activity (Table 1). In particular, calcium recovered the activity of HypBA2-C⌬893, but only partly recovered the activities of -C⌬1026 and -C⌬1049. Therefore, the F5/8 domain (aa 908 -1044) is likely to be the calcium-binding domain and related to the substrate recognition.
Previously, Hyp-linked arabinofuranosides in plants have been analyzed by alkaline hydrolysis and column chromatography. Our finding that ␤-L-arabinobiosidase has strict substrate specificity for Ara 3 -Hyp can be used to develop a simpler technique to detect the presence and study the role of ␤-L-arabinofuranosides in plant glycoproteins.