Structural and biochemical characterization of novel bacterial α-galactosidases belonging to glycoside hydrolase family 31.

Biochemical Journal Immediate Publication. Published on 05 May 2015 as manuscript BJ20150261

Structural and biochemical characterization of novel bacterial α-galactosidases belonging to glycoside hydrolase family 31 Takatsugu Miyazaki, Yuichi Ishizaki, Megumi Ichikawa, Atsushi Nishikawa, Takashi Tonozuka1

sc rip t

Department of Applied Biological Science, Tokyo University of Agriculture and Technology, 3-5-8 Saiwai-cho, Fuchu, Tokyo 183-8509, Japan 1

To whom correspondence should be addressed: Takashi Tonozuka, Department of Applied Biological

Science, Tokyo University of Agriculture and Technology, 3-5-8 Saiwai-cho, Fuchu, Tokyo 183-8509, Japan;

nu Ma pte d ce Ac

THIS IS NOT THE VERSION OF RECORD - see doi:10.1042/BJ20150261

Tel.: +81-42-367-5702; Fax: +81-42-367-5705; E-mail: [email protected]

1

Licenced copy. Copying is not permitted, except with prior permission and as allowed by law. © 2015 The Authors Journal compilation © 2015 Biochemical Society


ABSTRACT Glycoside hydrolase family 31 (GH31) proteins have been reportedly identified as exo-α-glycosidases with activity for α-glucosides and α-xylosides. We focused on a GH31 subfamily, which contains proteins with low sequence identity (24%) to the previously reported GH31 glycosidases, and characterized two

sc rip t

enzymes from Pedobacter heparinus and Pedobacter saltans, respectively. The enzymes unexpectedly exhibited α-galactosidase activity, but were not active on α-glucosides and α-xylosides. The crystal structures of one of the enzymes, PsGal31A, in unliganded form and in complexes with D-galactose or L-fucose, and the catalytic nucleophile mutant in unliganded form and in complex with p-nitrophenyl-α-D-galactopyranoside, were determined at 1.85–2.30 Å resolution. The overall structure of PsGal31A contains four domains, and the catalytic domain adopts a (β/α)8-barrel fold that resembles the structures of other GH31 enzymes. Two catalytic aspartic acid residues are structurally conserved in the enzymes, whereas most residues forming the

nu

loop that is not conserved in other reported GH31 enzymes; this loop is involved in its aglycone specificity and in binding L-fucose. Considering potential genes for α-L-fucosidases and carbohydrate-related proteins within the vicinity of Pedobacter Gal31, the identified Gal31 enzymes are likely to function in a novel sugar

Ma

degradation system. This is the first report of α-galactosidases which belong to GH31 family. Summary statement

We identified two bacterial enzymes as the first members that displayed α-galactosidase activity, and the crystal structures provided insights into their novel substrate specificity. This is the first report of

pte d

α-galactosidases which belong to GH31 family.

Short title: Structures and characterization of novel GH31 α-galactosidases Keywords: Crystal structure, GH31, Pedobacter, polysaccharide utilization, α-galactosidase, L-fucose Abbreviations: PUL, polysaccharide utilization locus; CAZy, carbohydrate-active enzyme; GH, glycoside

ce

hydrolase; PL, polysaccharide lyase; pNP, p-nitrophenyl; Gb3-β-MP, 4-methoxyphenyl β-globotrioside; αGalF, α-galactopyranosyl fluoride; SeMet, selenomethionine; PDB, Protein Data Bank; RMSD, root mean square deviation.

Ac


active site differ from those of GH31 α-glucosidases and α-xylosidases. PsGal31A forms a dimer via a unique

2



INTRODUCTION Bacteria have elaborate survival strategies that enable utilization of various energy sources under different conditions. They are able to artfully uptake and degrade carbohydrates that are produced by eukaryotes including fungi, plants, and animals [1–4]. Polysaccharide utilization loci (PUL) such as starch-,

sc rip t

cellulose-, and hemicellulose PULs encode carbohydrate-degrading enzymes, carbohydrate-binding proteins, and transporters [3,5–8]. Carbohydrate-degrading enzymes are remarkably diverse, and are categorized into many families in the carbohydrate-active enzyme (CAZy) database (http://www.cazy.org) [9]. The number of novel enzyme families and members has risen in recent years. The largest group is the glycoside hydrolases (GH), which are categorized in 133 families (some of them are deleted) [9]. The GH family includes characterized GHs and their uncharacterized homologs. In theory, the database facilitates function and substrate specificity predictions for uncharacterized proteins; in practice, the large number of potential GHs

nu

The GH31 family contains more than 3,000 proteins, and approximately 100 GH31 glycosidases have activity for α-glucosides and α-xylosides. One of the best-characterized enzymes is α-glucosidase (EC 3.2.1.20), which is involved in the hydrolysis of maltooligosaccharides produced by starch degradation of

Ma

α-amylases in various organisms [10–16]. Other glycoside hydrolases include α-glucosidase II (α-1,3-glucosidase; EC 3.2.1.84) [17], sucrase-isomaltase (EC 3.2.1.48 and EC 3.2.1.10) [18], and α-xylosidase (EC 3.2.1.177) [19–23]. GH31 enzymes hydrolyze oligosaccharides by the retaining mechanism, in which a pair of aspartic acid residues acts as catalysts, with some enzymes catalyzing transglycosylation [10]. Some bacterial GH31 enzymes have greater transglycosylation activity than hydrolysis activity, and are

pte d

called transglycosylases; these include 6-α-glucosyltransferase [24], 3-α-isomaltosyltransferase [24], and oligosaccharide α-1,4-glucosyltransferase (EC 2.4.1.161) [25]. α-Glucan lyase (EC 4.2.2.13) also belongs to the GH31 family, and cleaves the α-1,4-glucosidic C1-O bond by abstraction of the C2 proton and formation of the enol form of anhydrofructose [26].

The crystal structures of several GH31 enzymes have been determined. They share the following four major domains: the N-terminal super-β-sandwich domain, the catalytic (β/α)8-barrel domain, and two C-terminal β-sandwich domains [11–14,16,18,21,22,25,26]. Structural differences in enzyme active sites and

ce

the presence of additional subdomains result in diverse substrate specificities of GH31 enzymes. GH31 members are classified into the larger GH-D clan, together with GH27 and GH36, both of which primarily contain α-galactosidases (EC 3.2.1.22) [11,27]. The original members of clan GH-D were GH27 and GH36 [27], and their structural similarity and common catalytic machinery have been well documented [28].

Ac


that have low sequence identity with characterized GHs makes functional predictions difficult.

Bioinformatical and structural comparison approaches provide evidence for a distant evolutional relationship between GH31 and GH27/GH36 [11,29,30]. All the families, GH27, GH31, and GH36, share the (β/α)8-barrel catalytic domain and conserved catalytic residues involved in a retaining mechanism [11]. Pedobacter heparinus (formerly known as Flavobacterium heparinum) is a glycosaminoglycan-degrading Gram-negative soil bacterium, and its whole genome has been sequenced [31,32]. The whole genome of Pedobacter saltans also has been sequenced, and it is predicted to express polysaccharide-degrading enzymes 3



[33]. Two and five proteins classified as GH31s have been identified in the genomes of P. heparinus and P. saltans, respectively. The primary structures of two proteins, P. heparinus Phep_2697 (named PhGal31A) and P. saltans Pedsa_3617 (named PsGal31A), have 72% identity with each other, whereas they show low sequence identity (24%) with characterized GH31 enzymes. In the present study, we characterized PhGal31A and PsGal31A as the first GH31 α-galactosidases that do

sc rip t

not have activity for α-glucosides and α-xylosides. We determined the crystal structure of PsGal31A, and the results clarified its novel substrate specificity, which is different from other α-galactosidases. EXPERIMENTAL Materials and bacterial strains p-Nitrophenyl-α-D-glucopyranoside p-nitrophenyl-α-D-mannopyranoside

p-nitrophenyl-α-D-galactopyranoside

(pNPαGal),

(pNPαMan),

p-nitrophenyl-α-D-xylopyranoside

(pNPαXyl),

(pNPαFuc),

p-nitrophenyl-β-L-arabinopyranoside

(pNPβArap),

and

nu

p-nitrophenyl-N-acetyl-α-D-galactosaminide (pNPαGalNAc) were purchased from Sigma-Aldrich (St. Louis, MO, USA). Trehalose (Glcα1–α1Glc), kojibiose (Glcα1–2Glc), nigerose (Glcα1–3Glc), maltose (Glcα1– 4Glc), maltotriose (Glcα1–4Glcα1–4Glc), isomaltose (Glcα1–6Glc), and pullulan were kindly provided by

Ma

Hayashibara Biochemical Laboratories (Okayama, Japan). Melibiose (Galα1–6Glc) was purchased from Kanto Chemical Co. (Tokyo, Japan), 4-methoxyphenyl Galα1–4Glcβ1–4Glc (Gb3-β-MP) was from Tokyo Chemical Industry Co. (Tokyo, Japan), and α(1-3)-galactobiose (Galα1–3Gal) and blood group B antigen tetraose type 5 [Galα1-3(Fucα1-2)Galβ1–4Glc] were from Elicityl OligoTech (Crolles, France). α-D-Galactopyranosyl fluoride (αGalF) was synthesized as described by Henze et al. [34]. All other chemicals

pte d

were reagent grade and obtained from standard commercial sources. The strains Pedobacter heparinus NBRC 12017 (DSM 2366) and Pedobacter saltans NBRC 100064 (DSM 12145) were purchased from NITE Biological Resource Center (Chiba, Japan). Escherichia coli strains JM109 and BL21 (DE3) were used for DNA manipulation and protein expression, respectively. Cloning, expression, purification, and mutagenesis

The genes for Phep_2697 (named PhGal31A) and Pedsa_3617 (named PsGal31A), without the

ce

hypothetical N-terminal signal sequences, were amplified from P. heparinus and P. saltans, respectively, by colony-direct PCR using EmeraldAmp PCR Master Mix (Takara Bio Inc., Shiga, Japan) and the primers listed in Supplementary Table S1, ligated into the pGEM-T Easy vector (Promega, Madison, WI, USA), and sequenced. Two bases in the PsGal31A open reading frame differed from the published sequence (GenBank

Ac


p-nitrophenyl-α-L-fucopyranoside

(pNPαGlc),

CP002545.1) reported by Liolios et al. [33]; therefore, the sequence determined in our study was submitted to DDBJ/EMBL/GenBank databases with the accession No. LC019121. Signal sequences were predicted by using the SignalP server [35]. PhGal31A without the N-terminal signal sequence was ligated into pET-21a(+) (Novagen, Madison, WI, USA) using NdeI and HindIII restriction sites, which generated a recombinant protein containing a C-terminal His-tag (-KLAAALEHHHHHH). For subcloning into pET-28a(+) (Novagen), an NheI restriction site in the PhGal31A open reading frame was removed by silent mutagenesis using 4



QuikChange Site-Directed Mutagenesis Kit (Agilent Technologies, Santa Clara, CA, USA) with the primer pair listed in Supplementary Table S1. PsGal31A without the N-terminal signal sequence was ligated into pET-28a(+) using NheI and XhoI restriction sites, which generated a recombinant protein containing an N-terminal His-tag and a thrombin-cleavage site (MGSSHHHHHHSSGLVPRGSHM-). The residue numbering in this paper is based on the native protein sequences including N-terminal signal peptides.

sc rip t

Site-directed mutagenesis was performed using the QuikChange Site-Directed Mutagenesis Kit (Agilent Technologies) with the desired primers (Supplementary Table S1) and the PsGal31A expression plasmid as a template. All sequence analyses were performed using the ABI PRISM 3130 Genetic Analyzer (Applied Biosystems, Foster City, CA, USA).

E. coli BL21 (DE3) harboring the expression plasmids was grown at 37°C in 1 litre of Luria-Bertani medium containing 50 μg/ml ampicillin or kanamycin. When the culture reached an optical density of 0.6 measured at 600 nm, it was induced with isopropyl-β-D-thiogalactopyranoside at a final concentration of 0.1

nu

resuspended in 30 ml of 50 mM sodium phosphate buffer (pH 8.0) containing 20 mM imidazole and 300 mM sodium chloride, and then disrupted by sonication. The soluble fraction was applied onto a nickel (Ni2+) nitrilotriacetic acid (Ni-NTA) agarose (Qiagen, Hilden, Germany) column, and the recombinant proteins were

Ma

eluted with the same buffer containing 250 mM imidazole. The enzymes were dialyzed against 10 mM HEPES-NaOH buffer (pH 7.0) and stored at 4°C. Selenomethione (SeMet)-substituted PsGal31A was obtained by expression in E. coli B834 (DE3) cultured in LeMaster medium [36], and was purified in a manner identical to purification of the native protein. Protein purity was confirmed by SDS-PAGE. Protein concentration was determined by measuring the absorbance at 280 nm based on theoretical molar absorption

Enzymatic assays

pte d

coefficients calculated using the ExPASy ProtParam server (http://web.expasy.org/protparam/).

Hydrolytic activity for pNP glycosides was measured in 50 µl reactions containing 2.5 µM of purified enzyme, 1 mM pNP glycoside, and 100 mM sodium citrate buffer (pH 5.2 for PhGal31A or pH 4.6 for PsGal31A) at 40°C (PhGal31A) or 38°C (PsGal31A). After incubation for 30 min, the reactions were quenched by adding 100 µl of 1 M sodium carbonate solution. Released pNP was quantified by measuring the

ce

absorbance at 400 nm and using a molar extinction coefficient of 17,400 M-1 cm-1, which was determined with p-nitrophenol as a standard. When gluco-oligosaccharides (trehalose, kojibiose, nigerose, maltose, maltotriose, and isomaltose), sucrose, and melibiose were employed as substrates, reaction mixtures containing 2.5 µM of purified enzyme, 50 mM substrate, and 100 mM sodium citrate buffer (pH 5.2 for PhGal31A or pH 4.6 for

Ac


mM, and grown for another 20 h at 20°C (PhGal31A) or 25°C (PsGal31A). The cells were harvested and

PsGal31A) were incubated at the temperatures stated above for 30 min. The reactions were stopped by boiling for 3 min, and the amount of glucose liberated was measured using the glucose oxidase-peroxidase method with the Glucose C-II Test Kit (Wako Pure Chemicals, Osaka, Japan). Hydrolyses of Gb3-β-MP, α(1-3)-galactobiose, and blood group B antigen tetraose type 5 were monitored by thin-layer chromatography (TLC) using silica gel 60 glass plates (Merck, Darmstadt, Germany) with acetonitrile/water (4:1, v/v) as the developing solvent, and the reaction products were visualized by spraying the plates with 5% (v/v) sulfuric 5



acid-methanol solution, followed by heating at 110°C for 5 min. Soluble starch, dextran, pullulan, pectin (from citrus), polygalacturonan, and galactomannans (guar gum and locust bean gum) were used for substrate screening, and the amount of reducing sugar released was quantified using the Somogyi-Nelson method [37]. The effect of pH was measured at 30°C using McIlvaine buffer (sodium phosphate-citrate pH 3.0–8.0) and 1 mM pNPαGal as the substrate. The effect of temperature was assayed at 25–60°C using 100 mM sodium

sc rip t

citrate buffer (pH 5.2 for PhGal31A or pH 4.6 for PsGal31A). To test pH stability, the enzymes were incubated at 30°C for 30 min in 47 mM glycine-HCl (pH 2.0–3.0), sodium citrate (pH 3.0–6.0), sodium phosphate (pH 6.0–8.0), bicine-NaOH (pH 8.0–9.0), or glycine-NaOH (pH 9.0–11.5). To test thermal stability, the enzymes were incubated at 20–65°C in 100 mM sodium citrate buffer for 30 min. The remaining activity for pNPαGal was examined under the standard condition described above, except a final concentration of 70 mM sodium citrate buffer (pH 5.2 or 4.6) was used.

Transglycosylation was performed in reaction mixtures containing 20 mM pNPαGal, 100–500 mM

D-glucose,

D-mannose,

N-acetyl-D-glucosamine,

D-galactose,

D-xylose,

nu

reaction product was detected by TLC as described above. The following acceptor substrates were used: D-fructose,

N-acetyl-D-galactosamine,

L-arabinose,

D-glucuronic

acid,

L-rhamnose,

L-fucose,

D-galacturonic

acid,

Ma

N-acetylneuraminic acid, myo-inositol, sorbitol, L-serine, L-threonine, methanol, ethanol, and glycerol. Kinetic studies

The initial velocities of the hydrolytic reaction for pNPαGal were determined using 100 mM sodium citrate buffer (pH 4.6) and seven concentrations of pNPαGal (4–40 mM). The enzyme concentrations used were 246

pte d

nM [wild-type (WT)] and 123 nM (R175A). The initial velocities of the hydrolytic reaction for αGalF were determined using 100 mM sodium phosphate buffer (pH 7.0), 24.6 nM of PsGal31A (WT and R175A), and five concentrations of αGalF (0.5–10 mM). The amount of released fluoride ion was determined using a fluoride-selective electrode (perfectION comb F combination electrode, Mettler-Toledo, Columbus, OH, USA). All kinetic assays were performed at 30°C. Kinetic parameters were calculated by nonlinear regression analysis using KaleidaGraph (Synergy Software, Reading, PA, USA).

ce

Crystallization, data collection, structure determination, and refinement Before crystallization, purified proteins were concentrated to 5–20 mg/ml using Amicon Ultra 10K ultrafiltration devices (Millipore). Proteins were crystallized at 20°C using the hanging-drop vapor diffusion method, in which 1.0 µl of protein solution in 10 mM HEPES-NaOH buffer (pH 7.0) was mixed with an equal

Ac


acceptor substrates [10% (v/v) for alcohols], and 100 mM sodium citrate buffer (pH 4.6) at 30°C, and the

volume of a crystallization reservoir solution. Initial crystallization screening was performed using Crystal Screen, Crystal Screen 2, and PEG/Ion Screen kits (Hampton Research, Aliso Viejo, CA, USA). No PhGal31A crystal was obtained under the conditions tested, whereas PsGal31A crystals were grown under conditions containing polyethylene glycol 4,000, polyethylene glycol 10,000, or polyethylene glycol monomethyl ether 2,000 as precipitants. Well-diffracting PsGal31A crystals were obtained with crystallization solution containing 15–25% (v/v) polyethylene glycol monomethyl ether 2,000 and 100 mM HEPES-NaOH 6



buffer (pH 7.5–8.5) or 100 mM Tris-HCl buffer (pH 8.0–9.0). Mutant D365A and SeMet-PsGal31A crystals also were grown under the same conditions as those used for native crystals. The crystal of WT protein in complex with D-galactose was obtained by co-crystallization under the above condition with HEPES-NaOH buffer containing 50 mM D-galactose. The crystal of WT protein in complex with L-fucose was prepared by soaking with the reservoir solution (Tris-HCl buffer) containing 500 mM L-fucose for 10 min. The crystal of

sc rip t

D365A in complex with pNPαGal was prepared by soaking with the reservoir solution (HEPES-NaOH buffer) containing 20 mM pNPαGal for 3 min. All crystals were cryoprotected with the mother liquor supplemented with ethylene glycol at a final concentration of 25% (v/v), and then flash-frozen in liquid nitrogen.

Diffraction data were collected at PF BL5A, PF-AR NW12A, and PF-AR NE3A beamlines (Photon Factory, Tsukuba, Japan). All data were processed and scaled using HKL2000 [38]. Initial phases were calculated from the single-wavelength anomalous dispersion data set of the SeMet-PsGal31A crystal using the AutoSol program in the PHENIX suite [39]. The resulting coarse-scale model of SeMet-PsGal31A was

nu

the MOLREP program in the CCP4 suite [41]. The structures of mutants and complexes with ligands were solved by the molecular replacement method using MOLREP, with the native structure as a search model. Refinement was performed using REFMAC5 in the CCP4 suite [42], and manual adjustment and rebuilding of

Ma

the model were performed using COOT [43]. Solvent molecules were introduced using ARP/wARP. Structure validation was performed using MolProbity [44]. Data collection and refinement statistics are listed in Table 1. Analysis of protein-ligand interactions was performed using LIGPLOT [45] and COOT. Protein assembly was evaluated by the Protein Interfaces, Surfaces, and Assemblies (PISA) server (http://www.ebi.ac.uk/pdbe/pisa/) [46]. Figures were prepared using PyMOL (http://www.pymol.org/). The coordinates and structure factors

pte d

have been deposited in the Protein Data Bank (PDB) under the accession codes listed in Table 1. Molecular weight determination

The molecular weights of PhGal31A and PsGal31A were determined by gel filtration chromatography on a Superdex 200 HR 10/30 column (GE Healthcare, Chalfont St. Giles, UK) at a flow rate of 0.5 ml/min using the ÄKTApurifier chromatography system (GE Healthcare). The column was equilibrated with 20 mM sodium phosphate containing 0.2 M NaCl. Calibration was performed using blue dextran 2,000 (2,000 kDa),

ce

thyroglobulin (669 kDa), ferritin (440 kDa), catalase (232 kDa), lactate dehydrogenase (140 kDa), and albumin (66 kDa).

Sequence alignment and phylogenetics

Ac


constructed using ARP/wARP [40], and then was used for molecular replacement of the native data set using

Primary structure alignments and 3D-structure-based alignments were performed using the MUSCLE program [47] and the PROMALS3D server [48], respectively. Alignment figures were generated by ESPript 3.0 [49]. Phylogenetic analysis of GH31 was performed with the maximum-likelihood method using multiple alignment of the sequences listed in the CAZy database except for the proteins which are both partially sequenced and uncharacterized. Phylogenetic trees were constructed using MEGA6 [50]. To calculate amino acid conservation in Gal31 proteins, the ConSurf server [51] was used. 7



RESULTS AND DISCUSSION Phylogenetic analysis of GH31 proteins According to the CAZy database, GH31 contains more than 3,000 proteins with approximately 100 characterized enzymes, all of which have been identified as enzymes with activity for α-glucosides or

sc rip t

α-xylosides. The characterized enzymes include α-glucosidase (and maltase-glucoamylase), α-xylosidase, α-1,3-glucosidase, sucrase-isomaltase, α-glucan lyase, 3-α-isomaltosyltransferase, 6-α-glucosyltransferase, and oligosaccharide α-1,4-glucosyltransferase. To investigate the substrate specificity distribution of GH31 enzymes, the phylogenetic tree of GH31 proteins was generated using 1,251 GH31 protein sequences available in the CAZy database (Figure 1A). The phylogenetic tree facilitated predictions of substrate specificity for uncharacterized proteins that belong to the same clades as characterized proteins, and revealed several distinct clades of uncharacterized proteins that have low sequence identity (Figure 1A). To clarify

nu

(named Gal31 in the present study) was targeted. Gal31 proteins were found in bacteria belonging to the phyla Bacteroidetes and Firmicutes (Figure 1B). P. heparinus has been reported to express many unique carbohydrate-degrading enzymes derived from vertebrates, such as polysaccharide lyases with activity for

Ma

glycosaminoglycans [31,32]. Therefore, two Gal31 proteins, Phep_2697 (named PhGal31A) from P. heparinus and its ortholog Pedsa_3617 (named PsGal31A) from P. saltans, were cloned, expressed in E. coli, and characterized. These proteins have 72% sequence identity with each other, and 24% identity with structurally characterized GH31 enzymes.

pte d

General properties and substrate specificities of PhGal31A and PsGal31A The His-tagged PhGal31A and PsGal31A recombinants were successfully expressed in soluble form, and yielded approximately 40 mg per liter of culture. For initial substrate screening, we tested several pNP glycosides as substrates. Both enzymes displayed hydrolytic activity for pNPαGal (10.1±0.1 and 10.7±0.1 nmol/minmg for PhGal31A and PsGal31A at 30°C, respectively), but neither enzyme displayed hydrolytic activity for pNPαGlc and pNPαXyl, which were normally hydrolyzed by the characterized GH31 enzymes. No hydrolytic activity was observed for pNPαMan, pNPαFuc, pNPαGalNAc, and pNPβArap. Oligosaccharide

ce

hydrolysis was tested next, and neither enzyme hydrolyzed gluco-oligosaccharides containing any type of α-glucosidic linkage and sucrose, whereas trace activity was detected for melibiose (data not shown). Some GH31 enzymes acting on α-glucosidic linkages have been reported to prefer transglycosylation to hydrolysis [24,25], but PhGal31A and PsGal31A did not display transglycosylation activity when using

Ac


structure-function relationships of uncharacterized GH31 proteins, one of the uncharacterized protein clades

maltooligosaccharides as substrates (data not shown). Further substrate screening was performed with naturally occurring carbohydrates such as α(1-3)-galactosides and α(1-4)-galactosides, which were identified in blood group B antigen and glycosphingolipids, respectively. However, no hydrolysis product was detected by TLC for Gb3-β-MP, α(1-3)-galactobiose, and blood group B antigen tetraose type 5 (data not shown). The enzymes did not display hydrolytic activity for any polysaccharides (see Experimental) including galactomannans, which contain a β(1-4)-mannan main chain and α(1-6)-galactosyl side chains. These results 8



suggested that PhGal31A and PsGal31A were the first GH31 enzymes with activity for α-galactosides, and have different substrate specificity than that of the characterized α-galactosidases, which hydrolyze α(1-3)-, α(1-4)-, and α(1-6)-galactoside linkages [52–55]. The effects of pH and temperature on hydrolysis by PhGal31A and PsGal31A were assayed using pNPαGal as a substrate. PhGal31A and PsGal31A had the highest hydrolytic activity at pH 4.6 and pH 5.2,

sc rip t

respectively (Supplementary Figure S1A). The optimum temperatures for PhGal31A and PsGal31A hydrolysis were 38°C and 40°C, respectively (Supplementary Figure S1B). PhGal31A and PsGal31A were stable (90%) in pH ranges between 4.5–9.5 and 5.0–10.5, respectively, at 30°C for 30 min (Supplementary Figure S1C), and were stable (90%) up to 50°C and 30°C, respectively, during 30 min incubation (Supplementary Figure S1D). The enzymes had similar properties, except that PsGal31A thermostability was lower than that of PhGal31A. The effects of salts, metal ions, and organic solvents on enzyme activity were examined using PsGal31A (Supplementary Table S2). A reduction in hydrolytic activity was observed in

nu

glycerol at concentrations of 25% (v/v) did not inactivate the enzyme. The addition of 100 mM NaCl slightly enhanced hydrolysis, whereas 500 mM NaCl markedly reduced enzyme activity. No major effect was observed with addition of metal ions (Ca2+, Mg2+, Mn2+, Co2+, Ni2+, Cu2+, and Zn2+) and EDTA, suggesting

Ma

that the enzyme may not require metals for activity.

To clarify the substrate specificity of the Gal31 enzymes, kinetic studies were performed using PsGal31A. The Km value for pNPαGal was 30±3 mM, which was 15–143 times higher than that of other GH31 enzymes for pNP glycosides (Table 2). PsGal31A displayed very low activity (kcat/Km = 3.0 × 10-2 s-1mM-1) for pNPαGal compared with the activities of GH27 and GH36 α-galactosidases toward pNPαGal, melibiose, and

pte d

raffinose. Then, αGalF was synthesized and used as a substrate because its released aglycone was fluoride ion, which was smaller and expected to have less effect on aglycone specificity than pNP. The kcat and Km values of PsGal31A for αGalF were 37±2 s-1 and 1.6±0.2 mM, which are comparable with those of GH31 enzymes for glycosyl fluoride substrates (Table 2). These results suggest that PsGal31A is an α-galactosidase with strict aglycone specificity, which differs from that of other known α-galactosidases. Effects of monosaccharides on enzyme activity

ce

The mechanism of hydrolytic catalysis of GH31 glycosidases is the retaining mechanism, and some GH31 glycosidases display transglycosylation activity. PsGal31A transglycosylation activity was examined using pNPαGal as donor substrate and several different monosaccharides and other compounds with hydroxyl groups as acceptor substrates (see Experimental). No transglycosylation product was detected, whereas

Ac


reactions containing ethanol, 2-propanol, ethylene glycol, and dimethyl sulfoxide, whereas methanol and

inhibition of hydrolysis for pNPαGal was observed in reactions containing D-galactose or L-fucose by TLC analysis (data not shown). The effects of monosaccharides and derivatives on PsGal31A activity were quantitatively investigated. As expected, hydrolysis was strongly inhibited by D-galactose, and PsGal31A displayed 20% and 4.4% activity in the presence of 100 mM and 500 mM D-galactose, respectively (Supplementary Figure S2A). L-Fucose weakly inhibited the activity compared with that of D-galactose, and 83% and 47% activity was observed in the presence of 100 mM and 500 mM L-fucose, respectively, whereas 9



the other monosaccharides and their derivatives did not markedly inhibit enzyme activity. The inhibition pattern of L-fucose was estimated using two concentrations of 50 mM and 100 mM, and then L-fucose exhibited competitive inhibition against pNPαGal (Ki = 75 mM) (Supplementary Figure S2B). L-Fucose is a synonym for 6-deoxy-L-galactose, and the stereochemistries of D-galactose and L-fucose are inverse, but

sc rip t

these results suggested that PsGal31A might accept L-fucose in the active site. Overall structure of PsGal31A

The SeMet-PsGal31A crystal structure was solved at 2.39 Å resolution by the single-wavelength anomalous dispersion method. The crystal structure of native PsGal31A was determined by molecular replacement at 2.10 Å resolution. The crystal structures of the WT enzyme in complex with D-galactose (WT-Gal) and L-fucose (WT-Fuc), and the catalytic nucleophile mutant D365A (described below) in unliganded form (D365A) and in complex with pNPαGal (D365A-pNPαGal) were determined at 2.30, 1.85,

nu

contain one molecule in an asymmetric unit. The electron density (2Fo–Fc) maps for the protein contoured at 1 σ showed continuous density for all main chain atoms, except for residues 0–22 (for WT-Fuc) or 0–28 (for other structures) containing the N-terminal His-tag, and residues 538–563. PsGal31A contains an N-terminal

Ma

domain (N-domain, residues 29–207), a catalytic domain (A-domain, residues 208–531), a proximal C-terminal domain (C-domain, residues 532–628), and a distal C-terminal domain (D-domain, residues 629– 719) (Figure 2). The N-domain is a β-sandwich with two antiparallel β-sheets containing 12 β-strands (βN1– βN12) and two α-helices (αN1 and αN2). The A-domain adopts a (β/α)8-barrel fold, where a short region (310–333) with an α-helix inserts. The A-domain includes 10 α-helices (αA0–αA9 and αA; αA5 is broken)

pte d

and 8 β-strands (βA1–βA8). The C-domain contains a small β-sandwich fold with 8 β-strands (βC1–βC8). The D-domain adopts a β-sandwich fold with 6 β-strands (βD1–βD6). The root mean square deviations (RMSD) for Cα atoms among these structures are 0.109–0.241 Å2, indicating that the whole structures are almost identical. No metal ions were identified in the structures, in agreement with the biochemical study described above.

The whole structure of PsGal31A is similar to those of other GH31 members despite the low sequence identity (24%). Structural homology search was performed using the Dali Lite v. 3 server [56], and the

ce

highest Z score was observed with an α-xylosidase from Cellvibrio japonicus Ueda107 (CjXyl31A, PDB 2XVL, Z score = 37.9, sequence identity = 21%) [22], followed by an α-glucosidase from sugar beet (Beta vulgaris) (SBG, PDB 3WEO, Z score = 37.1, sequence identity = 19%) [16], and other structure-known GH31 enzymes (Supplementary Table S3). GH27 and GH36 α-galactosidases also were found in the Dali results, but

Ac


1.85, and 2.10 Å resolutions, respectively (Table 1). These crystals belong to the space group P43212, and

their Z scores (19.2) and sequence identities (14%) were lower than those of GH31 enzymes (Supplementary Table S3). The most remarkable differences between the tertiary structures of PsGal31A and other known GH31s are two inserted loops in the N-domain and C-domain of PsGal31A (Figure 2). The C-domain insert (residues 538–563) was disordered, whereas the electron density for the N-domain insert (loop-N, residues 161–186) was clear, and loop-N could be fully modeled. Loop-N is located between βN10 and βN11 and is conserved in 10



the ortholog PhGal31A (Figure 3A) and in other Gal31 proteins (Supplementary Figure S3). Loop-N is not found in GH31 α-glucosidases and α-xylosidases (Figure 3A), and CjXyl31A has a PA14 domain in the corresponding region [22]. Although there is one monomer in the asymmetric unit as described above, PsGal31A appears to form a dimer with the molecule related by the crystallographic 2-fold axis (Figure 3B), which is consistent with the gel filtration chromatography analysis (Figure 3C). The molecular mass of

sc rip t

PsGal31A was estimated as 140 kDa despite the theoretical molecular weight of 81.3 kDa, suggesting that the enzyme was dimeric in solution, similarly as PhGal31A (theoretical mass = 79.5 kDa; experimental mass by gel filtration = 130 kDa). In the dimer, loop-N extends toward the catalytic domain of the other subunit and forms many hydrogen bonds between loop-N and the counterpart. Analysis using the PISA server [46] supported the dimerization, and identified 24 hydrogen bonds and 6 salt bridges in the dimer interface (Supplementary Table S4). The total surface area of a monomer is 25,381 Å2, and the buried interface area is 2,372 Å2 (9.3% of the monomer surface). Twenty one residues per monomer are involved in these interactions,

nu

A-domain insert. A loop-N-deleted mutant named Δloop-N (residues 161–186 were deleted) was constructed and analyzed by gel filtration similarly as the WT enzyme, resulting in 90% dissociation of the dimer (Figure 3C). Several GH31 glycosidases have been reported to form oligomers, such as YicI, hexamer (trimer of

Ma

dimers) [21]; the hexamer MalA (dimer of trimers) [11]; and the dimer Ro-αG1 [13]. Ro-αG1 is the only structurally determined GH31 protein reported to be a dimer, and the enzyme dimerizes via an α-helical hairpin which inserts between the regions corresponding to βA4 and αA4 of PsGal31A. These results suggest that loop-N plays a major role in the dimerization, whose arrangement is quite different from that of the

pte d

Ro-αG1 dimer. Subsite –1 and catalytic residues

The electron density for β-galactose (named βGal –1) was clearly observed at subsite –1 (subsite nomenclatures are according to Davies et al. [57]) of WT-Gal, and βGal –1 was strictly recognized by the residues forming subsite –1 with hydrogen bonds (Figure 4A and B). Glu266, Tyr274, Lys363, Asp365, Arg418, Asp434, and Asn484 form hydrogen bonds with hydroxyl groups of βGal –1, whereas Trp486 makes van der Waals interactions between the O3, C3, C4, C5, C6, and O6 atoms of βGal –1. The latter interaction is

ce

not possible with glucose and xylose molecules due to their equatorial hydroxyl groups at the 4 position, in agreement with the result showing no activity for α-glucosides and α-xylosides. Superposition into the structures of NtMGAM and YicI indicated that Asp365 and Asp434 were structurally conserved with the catalytic nucleophile and acid/base, respectively, in other GH31 enzymes

Ac


and 7 of these residues are located in loop-N, and the others are included in N-domain, A-domain, and

(Figure 4C and D). The other residues were not well conserved with YicI and NtMGAM, except for Arg418, whose corresponding residues were fully conserved among GH31 enzymes (Supplementary Figure S4). The active site of PsGal31A was compared with that of an uncharacterized GH31 protein from Listeria monocytogenes, Lmo2446, which has 53% sequence homology to 3-α-isomaltosyltransferase from Sporosarcina globispora [24] and low homology (23%) to PsGal31A but is phylogenetically closer than the other structure-determined GH31 enzymes (Figure 1A). In addition to three residues (Asp365, Arg418, and 11



Asp434), Glu266, Trp305, and Lys363 were conserved between PsGal31A and Lmo2446 (Figure 4E), whereas the other residues were not conserved, and especially no residue corresponding to Tyr274 was found in Lmo2446. Considering that the enzyme did not display hydrolytic activity on pNPβArap, the interaction between Tyr274 and O6 atom of the galactosyl residue was likely to be important for substrate binding. There is no space which can accept additional group around βGal –1 except for the C1 positon. The subsite –1

sc rip t

residues of PsGal31A are well conserved among Gal31 proteins according to the sequence alignment (Supplementary Figure S3), suggesting that other Gal31 proteins are likely to be active on α-galactosides. The PsGal31A catalytic site was compared with those of GH27 and GH36 α-galactosidases, human GalA (hGalA, PDB 3HG5), and Lactobacillus acidophilus Mel36A (LaMel36A, PDB 2XN2), respectively (Figure 4F and G), which are classified into the same clan (GH-D) together with GH31 [11]. Despite low sequence identity (

Novel β-1,4-Mannanase Belonging to a New Glycoside Hydrolase Family in Aspergillus nidulans.

Crystal structure of a raw-starch-degrading bacterial α-amylase belonging to subfamily 37 of the glycoside hydrolase family GH13.

Diverse modes of galacto-specific carbohydrate recognition by a family 31 glycoside hydrolase from Clostridium perfringens.

Structural and biochemical analyses of glycoside hydrolase family 26 β-mannanase from a symbiotic protist of the termite Reticulitermes speratus.

Evolution and functional characterization of CAZymes belonging to subfamily 10 of glycoside hydrolase family 5 (GH5_10) in two species of phytophagous beetles.

Expression and crystallization of a bacterial glycoside hydrolase family 116 β-glucosidase from Thermoanaerobacterium xylanolyticum.

Crystal structure and characterization of the glycoside hydrolase family 62 α-L-arabinofuranosidase from Streptomyces coelicolor.

Structural elements responsible for the glucosidic linkage-selectivity of a glycoside hydrolase family 13 exo-glucosidase.

Structural and Functional Characterization of a Ruminal β-Glycosidase Defines a Novel Subfamily of Glycoside Hydrolase Family 3 with Permuted Domain Topology.

A novel β-xylosidase structure from Geobacillus thermoglucosidasius: the first crystal structure of a glycoside hydrolase family GH52 enzyme reveals unpredicted similarity to other glycoside hydrolase folds.

Structural analysis of Clostridium acetobutylicum ATCC 824 glycoside hydrolase from CAZy family GH105.

Active site and laminarin binding in glycoside hydrolase family 55.

Enzymatic characterization of a glycoside hydrolase family 5 subfamily 7 (GH5_7) mannanase from Arabidopsis thaliana.

Characterization of a highly thermostable glycoside hydrolase family 10 xylanase from Malbranchea cinnamomea.

Proteomic Analysis of a Novel Bacillus Jumbo Phage Revealing Glycoside Hydrolase As Structural Component.

Molecular characterization of a family 5 glycoside hydrolase suggests an induced-fit enzymatic mechanism.

A novel member of glycoside hydrolase family 30 subfamily 8 with altered substrate specificity.

Distinctive molecular and biochemical characteristics of a glycoside hydrolase family 20 β-N-acetylglucosaminidase and salt tolerance.

Structural and functional studies of the glycoside hydrolase family 3 β-glucosidase Cel3A from the moderately thermophilic fungus Rasamsonia emersonii.

Gene cloning and biochemical characterization of eryngase, a serine aminopeptidase of Pleurotus eryngii belonging to the family S9 peptidases.

N-glycosylation patterns in two α-l-arabinofuranosidases from Penicillium canescens belonging to the glycoside hydrolase families 51 and 54.

Biochemical Characterization of the Lactobacillus reuteri Glycoside Hydrolase Family 70 GTFB Type of 4,6-α-Glucanotransferase Enzymes That Synthesize Soluble Dietary Starch Fibers.

Structural Insights into the Carbohydrate Binding Ability of an α-(1→2) Branching Sucrase from Glycoside Hydrolase Family 70.

β-Glucuronidase from Lactobacillus brevis useful for baicalin hydrolysis belongs to glycoside hydrolase family 30.