Elucidation of the Molecular Basis for Arabinoxylan-Debranching Activity of a Thermostable Family GH62 α-l-Arabinofuranosidase from Streptomyces thermoviolaceus

Abstract

Xylan-debranching enzymes facilitate the complete hydrolysis of xylan and can be used to alter xylan chemistry. Here, the family GH62 α-l-arabinofuranosidase from Streptomyces thermoviolaceus (SthAbf62A) was shown to have a half-life of 60 min at 60°C and the ability to cleave α-1,3 l-arabinofuranose (l-Araf) from singly substituted xylopyranosyl (Xylp) backbone residues in wheat arabinoxylan; low levels of activity on arabinan as well as 4-nitrophenyl α-l-arabinofuranoside were also detected. After selective removal of α-1,3 l-Araf substituents from disubstituted Xylp residues present in wheat arabinoxylan, SthAbf62A could also cleave the remaining α-1,2 l-Araf substituents, confirming the ability of SthAbf62A to remove α-l-Araf residues that are (1→2) and (1→3) linked to monosubstituted β-d-Xylp sugars. Three-dimensional structures of SthAbf62A and its complex with xylotetraose and l-arabinose confirmed a five-bladed β-propeller fold and revealed a molecular Velcro in blade V between the β1 and β21 strands, a disulfide bond between Cys27 and Cys297, and a calcium ion coordinated in the central channel of the fold. The enzyme-arabinose complex structure further revealed a narrow and seemingly rigid l-arabinose binding pocket situated at the center of one side of the β propeller, which stabilized the arabinofuranosyl substituent through several hydrogen-bonding and hydrophobic interactions. The predicted catalytic amino acids were oriented toward this binding pocket, and the catalytic essentiality of Asp53 and Glu213 was confirmed by site-specific mutagenesis. Complex structures with xylotetraose revealed a shallow cleft for xylan backbone binding that is open at both ends and comprises multiple binding subsites above and flanking the l-arabinose binding pocket.

INTRODUCTION

Xylan is a major plant cell wall polysaccharide that has been harnessed to produce different chemicals, fuels, and materials; the bioactive properties of xylans are also valued in food applications (1–3). Xylan consists of a linear β-1,4-linked xylopyranosyl (β-d-Xylp) backbone, which, depending on the source, can be partially substituted at O-2 positions and/or O-3 positions with α-l-arabinofuranosyl (α-l-Araf) substituents and acetyl groups and at O-2 positions with α-d-glucuronic acid (GlcA) or 4-O-methyl GlcA (MeGlcA) (4–6). The O-5 position of some arabinose residues can be further esterified with p-coumaric or ferulic acid (7, 8). Whereas glucuronoxylan is the predominant form of xylan in hardwood trees, xylan from coniferous softwoods contains both arabinofuranosyl and MeGlcA substituents (9); xylans from cereals and grasses have varying arabinose and MeGlcA contents. Consistent with the molecular diversity of xylans, its complete enzymatic hydrolysis typically requires the concerted actions of several enzymes, including endoxylanases and β-xylosidases, which target glycosidic linkages within the xylan backbone, along with α-l-arabinofuranosidases, α-glucuronidases, acetyl xylan esterases, feruloyl esterases, as well as glucuronoyl esterases, which target branching substituents in different xylan types (10–12). Of these activities, arabinofuranosidases are represented in a particularly large number of carbohydrate-active enzyme (CAZy) families (http://www.cazy.org/).

Based on amino acid sequence, α-l-arabinofuranosidases (EC 3.2.1.55) have been classified into glycoside hydrolase (GH) families GH3, GH43, GH51, GH54, and GH62. Moreover, family GH1 includes exo-acting arabinofuranosidases active on p-nitrophenyl-α-arabinofuranoside (pNP-AraF) and α-1,5 arabino-oligosaccharides (13), family GH30 includes enzymes with 1,5-α-l-arabinobiose activities (14), and family GH93 includes exo-acting enzymes that release arabinobiose from the nonreducing end of α-1,5-l-arabinan (15). Several of these enzyme families are further grouped into glycoside hydrolase clans GH-A (GH1, GH30, and GH51), GH-E (GH93), and GH-F (GH43 and GH62), reflecting the structural diversity of α-l-arabinofuranosidases. For example, the catalytic domain of enzymes belonging to clan GH-A adopt a (β/α)₈ triosephosphate isomerase (TIM) barrel fold, as exemplified by the structure of the GH51 α-l-arabinofuranosidase from Geobacillus stearothermophilus T-6 (16). In contrast, clan GH-E enzymes adopt a six-bladed β-propeller architecture (15), GH43 enzymes in clan GH-F display a “non-Velcroed” five-bladed β-propeller arrangement (17, 18), and the catalytic domain of the GH54 α-l-arabinofuranosidase from Aspergillus kawachii adopts a β-sandwich fold (19).

α-l-Arabinofuranosidases also vary in terms of substrate preference and the linkages that they hydrolyze. For example, GH51 as well as GH54 enzymes include α-l-arabinofuranosidases able to remove branching α-l-Araf moieties from arabinan and xylans, although in most cases so far, enzymes from these families tend to prefer p-nitrophenyl-α-arabinofuranoside (pNP-AraF) over polymeric substrates (20–23). GH3 arabinofuranosidases include bifunctional enzymes displaying both α-l-arabinofuranosidase and β-d-xylopyranosidase activities with pNP substrates (24). In comparison, family GH43 includes enzymes with a broad range of activities, including enzymes that can hydrolyze pNP-AraF (25), endo-α-l-arabinanases (26, 27), exo-α-1,5-l-arabinanases (28), and bifunctional α-l-arabinofuranosidase/β-d-xylosidases (29). Family GH43 also includes arabinoxylan α-l-arabinofuranohydrolases (AXHs) that remove α-l-Araf residues that are (1→2) and (1→3) linked to monosubstituted β-d-Xylp (AXH-m2,3) (30) and that cleave α-l-Araf residues that are (1→3) linked to disubstituted β-d-Xylp (AXH-d3) (31).

Recently, the first three-dimensional (3-D) structures of GH62 enzymes were reported, one from Ustilago maydis (UmAbf62A), one from Podospora anserina (PaAbf62A), and one from Streptomyces coelicolor (ScAraf62A) (32, 33). These studies confirmed that GH62 enzymes adopt a five-bladed β-propeller fold common to GH43 enzymes (32). While those studies generated important insights into structural and functional characteristics of GH62 enzymes, there are comparatively few examples of detailed biochemical and structural analyses of this GH family. Given the high activity of GH62 enzymes on polymeric arabinoxylan over low-molecular-weight substrates, including pNP-AraF (32–38), GH62 α-l-arabinofuranosidases are particularly relevant candidates for fine-tuning xylan chemistry as well as enabling the complete saccharification of xylan to sugars.

Accordingly, to deepen our understanding of GH62 enzymes, we report the biochemical and structural characterization of a thermostable bacterial GH62 α-l-arabinofuranosidase (SthAbf62A) from Streptomyces thermoviolaceus subsp. pingens Henssen, which has 67%, 58%, and 31% amino acid sequence identities to ScAraf62A, UmAbf62A, and PaAbf62A, respectively. Consistent with the first report of SthAbf62A by Tsujibo et al. (37), SthAbf62A characterized here released reducing sugars from polymeric arabinoxylan. This analysis further demonstrated that SthAbf62A can cleave α-l-Araf substituents that are (1→2) and (1→3) linked to monosubstituted β-d-Xylp in arabinoxylan and can also hydrolyze nonreducing terminal α-1,5 linkages in debranched arabinan as well as pNP-arabinofuranoside albeit with much lower specific activities. Structural analysis of SthAbf62A complexes with l-arabinose and xylotetraose revealed a Velcroed five-bladed-β-propeller and provided insight into the molecular basis of substrate selectivity for this enzyme.

MATERIALS AND METHODS

Chemicals.

Pfx polymerase was purchased from Invitrogen. The Qiagen stool kit, Qiaex II gel extraction kit, and QIAquick PCR purification kit were purchased from Qiagen. Wheat arabinoxylan (high viscosity [WAX-HV] and low viscosity [WAX-LV]), rye flour arabinoxylan (RAX), arabinan from sugar beet, debranched arabinan, rhamnogalacturonan (soy bean), larch arabinogalactan, and α-l-arabinofuranosidase from Bifidobacterium sp. (E-AFAM2; AXH-d3) were purchased from Megazyme, whereas oat spelt xylan (OSX), birchwood xylan, beechwood xylan, and p-nitrophenyl glycosides were obtained from Sigma-Aldrich. Soluble and insoluble fractions of OSX were prepared as previously reported (39). Streptomyces thermoviolaceus subsp. pingens Henssen (ATCC 19283) was purchased from the American Type Culture Collection (ATCC). All other chemicals were analytical grade and were obtained from Sigma-Aldrich or Fisher Scientific.

DNA manipulations.

Streptomyces thermoviolaceus subsp. pingens Henssen was grown in 5 ml Luria-Bertani medium at 37°C for 4 days. Genomic DNA was extracted by using the Qiagen stool kit according to the manufacturer's instructions and including a 5-min incubation at 95°C, followed by an additional bead-beating step using Garnet bead tubes (0.70 mm; Mo Bio). The gene encoding the mature form of SthAbf62A (GenBank accession number KF958299) without a signal peptide (amino acid [aa] residues 1 to 35) was amplified by using the forward and reverse primers shown in Table S1 in the supplemental material and Pfx DNA polymerase (Invitrogen). The infusion cloning kit from Clontech was used to transfer purified PCR products to the p15Tv-L expression plasmid (GenBank accession number EF456736), generating p15Tv-L_SthAbf62A. Site-specific mutagenesis was carried out according to a modified QuikChange (Stratagene) method that uses partially overlapping primers (see Table S1 in the supplemental material) (40). All constructs were verified by DNA sequencing at the Center of Applied Genomics at the SickKids Hospital in Toronto, Ontario, Canada.

Purification of wild-type and mutant α-l-arabinofuranosidases.

Recombinant Escherichia coli BL21(λDE3) Codon Plus harboring p15Tv-L_SthAbf62A was propagated at 37°C in 1 liter of Luria-Bertani medium supplemented with 0.5 M d-sorbitol, 2.5 mM glycine betaine, 34 μg/ml chloramphenicol, and 100 μg/ml ampicillin. At an optical density at 600 nm (OD₆₀₀) of 0.6, the cultivation temperature was reduced to 15°C, and recombinant protein expression was induced with 0.5 mM isopropyl β-d-thiogalactopyranoside.

The next day, induced cultures were harvested by centrifugation at 6,000 × g for 10 min. Cell pellets (approximately 4.5 g [fresh weight]) were suspended in binding buffer (300 mM NaCl, 50 mM HEPES [pH 7.0], 5% glycerol, 5 mM imidazole) and disrupted by sonication. Cell debris was removed by centrifugation (17,500 × g for 20 min), and supernatants were passed through a 0.45-μm filter before being incubated with a 5.0-ml bed volume of Ni-NTA (Ni²⁺-nitrilotriacetate) resin (Qiagen) for 45 min at 4°C. Resin samples were then washed with 200 ml of washing buffer (300 mM NaCl, 50 mM HEPES [pH 7.0], 5% [vol/vol] glycerol, 50 mM imidazole), and bound protein was eluted with approximately 30 ml of elution buffer (300 mM NaCl, 50 mM HEPES [pH 7.0], 5% [vol/vol] glycerol, 250 mM imidazole). Active fractions were pooled, exchanged to 20 mM HEPES buffer (pH 7.0) using a Bio-Gel P10 column, and then further purified by anion chromatography using a Uno-Q column (5.0 ml) from Bio-Rad. A Bioshop Duoflow system (Bio-Rad, Canada) was used to pass 20 mM HEPES buffer (pH 7.0) with a NaCl gradient from 0 to 0.5 M through the Uno-Q column over 20 column volumes, and wild-type SthAbf62A as well as mutant enzymes were eluted with 0.15 M NaCl. A Bio-Gel P10 column was then used to transfer the purified protein to 20 mM HEPES buffer (pH 7.0) containing 300 mM NaCl, and centrifugal filter units (molecular weight cutoff, 10,000; Millipore) were used for concentrating protein. Protein aliquots were flash-frozen in liquid nitrogen and then stored at −80°C.

Protein concentrations were determined by using the Bradford assay (41) and bovine serum albumin as a standard. SDS-PAGE was performed, and gels were stained with Coomassie blue according to established procedures.

Enzyme activity assays.

Hydrolysis of arabinoxylans was monitored by measuring released reducing sugars using the Nelson-Smogyi procedure (42). The standard assay solution contained 0.5% (wt/vol) (5.0 mg/ml) polysaccharide in 0.5 ml of 100 mM HEPES buffer (pH 7.0). The reaction was initiated by the addition of an amount of enzyme determined to release products in a linear relation to time when incubated at 45°C for up to 30 min, which were also the temperature and incubation time used to measure enzyme activities. One unit of activity was defined as the amount of enzyme releasing 1 μmol l-(+)-arabinose equivalent per minute. l-(+)-Arabinose was used to generate a standard curve (0.05 to 0.6 mg/ml).

SthAbf62A activity on 4-nitrophenyl (pNP) glycosides was determined by measuring pNP release. The typical reaction mixture contained 2.0 mM substrate in 0.30 ml of 100 mM HEPES buffer (pH 7.0). After the reaction was initiated by the addition of enzyme, the reaction mixture was incubated for 20 min at 45°C. The reaction was terminated by adding 0.30 ml of 4% (wt/vol) Na₂CO₃ to the mixture. The amount of pNP formed was measured by the A₄₀₅ (extinction coefficient ε [405 nm] = 17,600 M⁻¹ cm⁻¹), and pNP was used to generate a standard curve. As also indicated above, the enzyme added to each assay mixture was optimized to measure initial reaction rates. All enzyme assays were carried out in triplicate.

Kinetic parameters were determined by using 0.1 mM to 25 mM pNP-AraF; apparent kinetic parameters were also obtained with 0.05% to 1.5% (0.5 to 15 mg/ml) WAX-HV and 0.1% to 15% (1.0 to 150 mg/ml) sugar beet arabinan. Initial rates were obtained by measuring reaction products formed after 30 min at 45°C at pH 7.0, as described above for the standard assay. Kinetic parameters were then calculated by using the Michaelis-Menten equation or the conventional substrate inhibition equation (see below) using GraphPad Prism 5.0 (GraphPad Software):

$$v=\frac{v_{\max \hspace{0.17em}}\cdot [S]}{K_{m(\mathrm{a}\mathrm{p}\mathrm{p})}+[S]\cdot (1+\frac{[S]}{K_{si(\mathrm{a}\mathrm{p}\mathrm{p})}})}$$

where v is the initial velocity at various substrate concentrations and K_m(app) is the apparent Michaelis constant for the substrate. K_si(app) represents the apparent dissociation constant for the inhibitory substrate-enzyme-substrate (SES) ternary complex (43).

Optimum reaction conditions, enzyme stability, and effects of divalent ions.

All enzyme assays were performed by using the standard activity assay with 0.5% (wt/vol) WAX-HV with incubation at 45°C for 30 min; reaction products were then detected by using the Nelson-Smogyi method. The effect of pH on SthAbf62A activity was determined by performing the activity assay at pH 3.5 to 10.5 with increments of 0.5 pH units; a universal buffer (100 mM acetic acid, 100 mM boric acid, and 100 mM phosphoric acid, adjusted to the target pH by using a sodium hydroxide solution) was used for this analysis. The effect of reaction temperature on enzyme activity was examined by performing the standard assay at temperatures ranging from 5°C to 85°C. To determine pH and temperature stability, the enzyme was preincubated in universal buffer (pH 4 to 10.5) at 4°C for 24 h, or in 100 mM HEPES buffer (pH 7.0) at 25°C to 80°C for 60 min, before being adjusted to standard assay conditions to measure residual SthAbf62A activity. The effect of metal ions was determined by adding 2.0 mM different metal chloride salts to the standard activity assay mixture.

High-pressure anion-exchange chromatography detection of reaction products.

To confirm the selective hydrolysis of arabinose substituents, products of WAX-HV and sugar beet arabinan from treatment with SthAraf62A were analyzed by using a high-pressure anion-exchange chromatography (HPAEC) instrument (ICS-5000⁺ DC; Dionex) equipped with a pulsed amperometric detector (PAD), a Carbopac PA1 column (250 by 2 mm; Dionex), and a Carbopac PA20 guard column (30 by 3 mm; Dionex) (44). Reaction mixtures (0.5 ml) contained 0.5% (wt/vol) WAX-HV or arabinan and SthAbf62A (2.5 μg for WAX-HV and 12.0 μg for arabinan), as described above for the standard activity assay. After 15 min, 30 min, and 45 min of incubation, reactions were stopped by boiling for 10 min; 100-μl samples were then mixed with 2 volumes of acetone to precipitate polymeric arabinoxylan and centrifuged at 13,000 × g for 10 min at 4°C. Supernatants containing the liberated sugars were transferred into a clean tube, and the acetone was removed by using a Speedvac instrument (Thermo). The remaining supernatant was lyophilized, dissolved in 100 μl ultrapure water (MilliQ), and then filtered through 0.22-μm centrifuge tube filters (Corning). Samples (10 μl) were injected onto a Carbopac PA1 column kept at 30°C and eluted with 50 mM NaOH at 0.25 ml/min for >30 min. l-Arabinose and d-xylose (40 μg/ml) were used as standards.

Preparation of wheat arabinoxylan with singly substituted l-arabinose.

Arabinoxylan with singly substituted α-1,2- and 1,3-arabinose was generated by selective removal of α-1,3-arabinose from doubly substituted xylose residues using arabinofuranosidase from Bifidobacterium sp. (AXH-d3) (36). Briefly, 0.2% (wt/vol) WAX-HV was treated with 5 U of AXH-d3 in 50 ml 20 mM potassium-phosphate buffer (pH 6.0) at 40°C for 24 h; an additional 2 U of AXH-d3 was then added, and the reaction mixture was incubated for another 24 h to ensure near-complete removal of the α-1,3-arabinose from doubly substituted positions. After boiling for 10 min, arabinoxylan containing singly substituted α-1,2- and 1,3-arabinose was precipitated by the addition of 2 volumes of 95% ethanol followed by centrifugation (15,000 × g for 10 min). The pellet was dissolved in 50 ml water and again precipitated with 95% ethanol. The precipitate was then dissolved in 20 ml water and lyophilized.

¹H NMR analysis of the mode of action of SthAbf62A toward arabinoxylan.

To qualitatively evaluate the regioselectivity of SthAbf62A on arabinoxylan, one-dimensional ¹H nuclear magnetic resonance (NMR) spectra were collected for intact and AXH-d3-treated WAX-HV before and after treatment with SthAbf62A. Briefly, 1% (wt/vol) arabinoxylan (intact or singly substituted) was digested for 2 h or 24 h at 45°C by using 3.0 μg of SthAbf62A in a 0.6-ml reaction solution containing 100 mM HEPES buffer (pH 7.0). Reactions were terminated by boiling for 10 min. The treated arabinoxylan was then precipitated by the addition of 1.2 ml of 95% ethanol and separated by centrifugation (15,000 × g for 10 min). The pellet was suspended in 0.6 ml water and again precipitated with 95% ethanol. To overcome the precipitation of arabinoxylan following treatment with arabinofuranosidases (45, 46), a GH11 endoxylanase from Thermobifida fusca, TfxA (47), was used to hydrolyze the debranched arabinoxylan into arabino-xylooligosaccharides (48). The pellet was suspended in 0.6 ml of 10 mM 3-(N-morpholino)propanesulfonic acid (MOPS) buffer (pH 7.0) and then treated with 1.2 μg TfxA at 50°C. After 24 h, an additional 1.2 μg TfxA was added, and the reaction mixture was incubated for another 24 h. Reaction mixtures were subsequently boiled for 15 min to inactivate the endoxylanase and then lyophilized and dissolved in 0.6 ml D₂O two times before NMR analysis. ¹H NMR spectra were obtained at 25°C in 5.0-mm NMR tubes (Norell) by using an Agilent DD2 600-MHz spectrometer with a scan number of 16 and a relaxation delay of 10 s. The data were recorded by using VnmrJ 4.0 (Agilent) and processed by using MestReNova 8.1.0 software (Mestrelab Research). The methylene protons at position 2 of MOPS (2.035 ppm) were used as the internal reference. The NMR signal assignment was based on data from previous reports (36, 49).

Protein crystallization, data collection, and structure determination.

SthAbf62A was concentrated to 18 mg/ml in 10 mM HEPES (pH 7.5) containing 500 mM NaCl. The crystallization screen was performed by mixing 0.5 μl of protein with 0.5 μl of reservoir buffer using a Mosquito crystallization robot. The SthAbf62A crystals were obtained by using a sitting-drop vapor diffusion technique at 25°C with a solution containing 0.1 M Bis-Tris propane (pH 6.5) and 20% polyethylene glycol monomethyl ether (5 K). For cocrystallization, SthAbf62A (14 mg/ml) was first incubated overnight at 4°C with xylotetraose and l-arabinose at final concentrations of 5 mg/ml and 200 mM, respectively. The cocrystallization reservoir buffer for xylotetraose contained 0.2 M sodium formate and 20% polyethylene glycol 3350, and that for l-arabinose contained 0.1 M HEPES (pH 7.5), 0.2 M ammonium acetate, and 25% polyethylene glycol 3350. The crystals were grown at 25°C by using the sitting-drop vapor diffusion technique. The crystals were then flash-frozen in a 100-K liquid nitrogen stream, using Paratone-N oil as a cryoprotectant. Diffraction data for SthAbf62A apoenzyme were collected at 100 K on a Rigaku HF-007 rotating anode with a Rigaku R-AXIS IV++ detector. Diffractor data for SthAbf62A-arabinose and SthAbf62A-xylotetraose complexes were collected at the Argonne National Laboratory Structural Biology Center at Advanced Photon Source beamline 19-ID with an ADSC Quantum 315r detector. All diffraction data were processed with HKL3000 (50). The SthAbf62A apoenzyme structure was solved by molecular replacement using the structure of a family GH62 α-l-arabinofuranosidase from Scytalidium thermophilum (our unpublished data) as the search model and PHENIX.phaser (51). All models were refined by using PHENIX and Coot (52) with 5% of observed reflections set aside for R_free calculations. Ligands were built into residual positive F_o − F_c density after building the protein atoms. All geometries were verified with PHENIX.refine and the RCSB PDB Validation server. The conformation of l-arabinose and nucleophilic water was verified by calculation of a simulated annealing omit map, where these atoms were removed, followed by Cartesian simulated annealing using PHENIX.refine (default parameters). Average B-factor and bond angle/bond length root mean square deviation (RMSD) values were calculated by using PHENIX. Table 4 lists data collection and final model refinement statistics. Structure figures were prepared by using the PyMol molecular graphics system (2002; DeLano Scientific, San Carlos, CA, USA).

TABLE 4.

X-ray diffraction data collection and refinement statistics

Parameter	Value(s) for liganda
	Apoenzyme	Xylotetraose	l-α-Arabinose
PDB accession no.	4O8N	4O8P	4O8O
Data collection statistics
Space group	P212121	P212121	P212121
Cell dimensions
a, b, c (Å)	59.81, 65.68, 84.95	42.49, 64.64, 121.30	60.05, 65.74, 85.20
α, β, γ (°)	90, 90, 90	90, 90, 90	90, 90, 90
Resolution (Å) (range)	25.00–1.65	25.00–1.56	18.00–1.18
Rmergeb	0.081 (0.587)	0.086 (0.517)	0.050 (0.508)
I/σ(I)	29.96 (4.0)	21.77 (3.4)	31.45 (3.2)
Completeness (%)	96.2 (92.7)	99.6 (99.5)	98.1 (96.1)
Redundancy	5.9 (6.0)	4.6 (4.7)	4.9 (4.8)
Refinement statistics
Resolution (Å)	24.45–1.65	23.68–1.56	17.47–1.21
No. of reflections	38657	48371	101459
R-factor/free R-factorc	12.7/16.3 (18.0/24.0)	13.0/15.3 (17.2/19.9)	13.9/15.6 (19.6/21.67)
No. of refined atoms
Protein	2,391	2,444	2,405
Ca2+	1	1	1
Ligand	NA d	35	10
Solvent	39	39	0
Water	540	474	516
B-factor
Protein	15.3	13.0	14.3
Ca2+	10.8	6.1	7.8
Ligand	NA	19.8	15.0
Solvent	35.9	53.2	NA
Water	33.1	31.6	33.2
RMSD
Bond lengths (Å)	0.012	0.019	0.012
Bond angles (°)	1.333	1.808	1.384
Ramachandran plot (%)
Most favored regions	84.9	87.4	85.8
Additionally favored regions	15.1	12.6	14.2
Generously favored regions	0	0	0
Disallowed regions	0	0	0

^aValues in parentheses represent high-resolution shells (4O8N = 1.65 to 1.68 Å, 4O8P = 1.56 to 1.59 Å, 4O8O = 1.18 to 1.20 Å for data collection and 1.21 to 1.24 Å for refinement).

^bR_merge = (Σ_hkl Σ_j |I_hkl,j − 〈I_hkl〉|)/(Σ_hkl Σ_j I_hkl,j), where 〈I_hkl〉 is the average of symmetry-related observations of a unique relection.

^cR-factor = [Σ_hkl |F_hkl (obs) − F_hkl (calc)|]/[Σ_hkl F_hkl (obs)], where obs = observed and calc = calculated; free R-factor calculated with 5% reflections set aside.

^dNA, not applicable.

Protein structure accession numbers.

Structures were deposited in the Protein Data Bank under accession numbers 4O8N (SthAbf62A), 4O8O (SthAbf62A complexed with l-arabinose), and 4O8P (SthAbf62A complexed with xylotetraose).

RESULTS AND DISCUSSION

General enzyme properties of SthAbf62A.

Recombinant SthAbf62A was functionally expressed in E. coli with an N-terminal His₆ tag and purified to homogeneity, with a typical protein yield of ∼12 mg protein per liter of bacterial culture. The molecular mass of recombinant SthAbf62A estimated by SDS-PAGE was 36.5 kDa (see Fig. S1 in the supplemental material), which is consistent with the predicted molecular mass (38.3 kDa). The optimal pH of SthAbf62A activity on wheat arabinoxylan was approximately pH 7.0 (Fig. 1A), which is 2.0 pH units higher than the one previously reported for the identical α-l-arabinofuranosidase from Streptomyces thermoviolaceus OPC-520 (GenBank accession number AB110643.1) (37). Although initially surprising, this difference in pH optima could be explained by the comparatively long incubation time (2 h) used in previous studies to measure the pH optimum together with the pH stability of SthAbf62A (see below). For comparison, reported pH optima for other characterized GH62 α-l-arabinofuranosidases are pH 6.0 for SliAraf62 from Streptomyces lividans (34), pH 5.5 for ScAraf62A from S. coelicolor (33), pH 5.0 for PchAraf62 from Penicillium chrysogenum (36), pH 7.0 for CcAraf62A from Coprinopsis cinerea (38), pH 5.0 for UmAbf62A from Ustilago maydis, and pH 5.0 for PaAbf62A from Podospora anserina (32).

FIG 1.

Effect of pH and temperature on SthAbf62A activity and stability. (A) Effect of pH on enzyme activity; (B) pH stability of purified SthAbf62A; (C) effect of temperature on SthAbf62A activity; (D) thermostability of purified SthAbf62A. Activity assays were performed with 0.5% wheat arabinoxylan (high viscosity) and 0.4 μg SthAbf62A.

Stability studies showed that SthAbf62A retained >70% activity after 24 h at 4°C between pH 6.0 and pH 9.5 (Fig. 1B). In comparison, CcAraf62A retains 80% activity after 1 h at 20°C between pH 7.0 and pH 8.5 (38), and ScAraf62A is stable for 1 h at 35°C between pH 7.0 and pH 9.0 (33). Furthermore, SthAbf62A showed the highest activity at 55°C in a 30-min enzyme assay (Fig. 1C) and retained >70% and 45% of activity after 60 min at 55°C and 60°C, respectively (Fig. 1D). In contrast, PchAraf62, SliAraf62, and ScAraf62A lost all activity after 60 min at 60°C (33, 34, 36), and the optimal reaction temperatures for UmAbf62A and PaAbf62A are 37°C and 55°C, respectively (32, 53).

Consistent with data from previous reports (37), the presence of 2.0 mM Zn²⁺, Cd²⁺, Hg²⁺, and Cu²⁺ reduced sugar release from WAX-HV by SthAbf62A by >70%, while 2.0 mM Ca²⁺ and Mg²⁺ activated SthAbf62A by at least 60% (see Fig. S2 in the supplemental material). However, addition of 2.0 mM EDTA did not significantly affect SthAbf62A activity, suggesting that the enzyme does not strictly depend on ion supplementation despite the coordinated calcium ion observed in the SthAbf62A structure (see below).

Substrate specificity of SthAbf62A.

SthAbf62A activity was highest on WAX-HV and RAX, followed by water-soluble OSX and WAX-LV (Table 1). This result is consistent with the considerably higher arabinose-to-xylose ratio in WAX-HV and RAX (ranging from 0.5 to 0.7) (54, 55) than in WAX-LV (approximately 0.3) (54) and OSX (approximately 0.12) (56, 57). Notably, however, when wheat, rye, and water-soluble OSX were normalized to the arabinose content, approximately 8% (WAX-HV), 6% (WAX-LV and RAX), and 17% (OSX) of the total arabinose present was released after 30 min by 0.4 μg SthAbf62A under standard assay conditions. This suggests that arabinose in water-soluble OSX is comparatively accessible to hydrolysis by SthAbf62A, which likely reflects the particular solubility of this OSX preparation. More interestingly, similar extents of arabinose release from wheat and rye arabinoxylan after 30 min suggest that the differences in the occurrences of α-l-Araf (1→3) and α-l-Araf (1→2) substituents in WAX-HV, WAX-LV, and RAW (54) do not significantly affect initial rates of SthAbf62A activity.

TABLE 1.

Activity of SthAbf62A on polymeric xylans and pNP synthetic substrates^b

Substrate (μg SthAbf62A)a	Mean sp act (μmol product/min/μmol enzyme) ± SD	Relative activity (%)
Polymeric substrates
WAX-HV (0.4)	1,535 ± 92	100.0
WAX-LV (0.4)	930 ± 55	60.6
RAX (0.4)	1,533 ± 118	99.9
Soluble OSX (0.4)	1,082 ± 66	70.5
Insoluble OSX (24)	80 ± 23	5.2
Arabinan from sugar beet (24)	47 ± 6	3.0
Debranched arabinan (60)	1 ± 0.4	0.07
Rhamnogalacturonan (60)	0.4 ± 0.1	0.02
pNP synthetic substrates
pNP-α-l-arabinofuranoside (2)	51 ± 2	100
pNP-α-l-arabinopyranoside (78)	0.3 ± 0.02	0.5
pNP-β-d-xylopyranoside (224)	0.02 ± 0.003	0.05

^aAmounts of purified SthAbf62A added to each reaction mixture to measure initial reaction velocities.

^bActivities obtained with WAX-HV and pNP-α-l-arabinofuranoside were taken as 100% in the respective comparisons. No activity was detected with birchwood xylan, beechwood xylan, apple pectin, larch arabinogalactan, pNP-β-l-arabinopyranoside, pNP-α-d-mannopyranoside, pNP-α-d-galactopyranoside, pNP-α-l-fucopyranoside, and pNP-α-d-glucopyranoside. All assays were performed under standard assay conditions.

Although not detected during the characterization of α-l-arabinofuranosidase from Streptomyces thermoviolaceus OPC-520 (37), SthAbf62A also released arabinose from sugar beet arabinan, which consists of an α-1,5-linked l-arabinofuranose backbone with ca. 60% α-1,3-linked l-arabinofuranose substitutions. HPAEC analyses further confirmed that l-arabinose was the only monosaccharide released from arabinoxylan and arabinan (see Fig. S3 in the supplemental material). Given the polydispersity of the arabinoxylan and arabinan substrates, molar values of these substrates could not be accurately calculated, and thus, only apparent kinetic parameters could be obtained to evaluate SthAbf62A activity in greater detail. These analyses revealed that the apparent catalytic efficiency (k_cat/K_m) of SthAbf62A was nearly 70 times higher on wheat arabinoxylan than on sugar beet arabinan (Table 2). At the same time, the kinetic data were best explained by the conventional substrate inhibition model (see the equation above), and apparent substrate inhibition constants [K_si(app)] were lower for arabinoxylan than for arabinan (Table 2; see also Fig. S4 in the supplemental material). Since sample viscosities did not perceptibly differ at maximum substrate concentrations, the lower K_si(app) value for arabinoxylan suggests preferred binding by SthAbf62A of xylan over arabinan backbone structures.

TABLE 2.

Kinetics of SthAbf62A on WAX-HV and sugar beet arabinan

Substrate (μg SthAbf62A)a	Mean kcat(app) (s−1) ± SD	Mean Km(app) (mg/ml) ± SD	Mean Ksi(app) (mg/ml) ± SD	kcat/Km (s−1 · mg−1 · ml)
WAX-HV (0.5)b	180 ± 65	12 ± 5	1.5 ± 0.5	15
Sugar beet arabinan (4)b	6 ± 0.5	32 ± 4.3	106 ± 18	0.2

^aAmounts of purified SthAbf62A added to each reaction mixture to measure initial reaction velocities.

^bThe conventional substrate inhibition equation (see the text) was used for data fitting.

Low but detectable rhamnogalacturonan, debranched arabinan, pNP-α-l-arabinofuranoside, and pNP-α-l-arabinopyranoside activities were also observed (Table 1). Notably, the catalytic efficiency (k_cat/K_m) of SthAbf62A on pNP-α-l-arabinofuranoside was ∼330 times higher than on pNP-α-l-arabinopyranoside, which was explained mainly by differences in k_cat values and reveals the high preference for furanose forms of α-l-arabinose (Table 3). HPAEC analyses confirmed that l-arabinose was the only product released from debranched α-1,5-linked arabinan (data not shown), and kinetic analyses determined that the k_cat values for arabinan and pNP-α-l-arabinofuranoside were 30 and 140 times lower than the apparent k_cat value for WAX-HV, respectively (Tables 2 and 3). The clear preference of SthAbf62A for polymeric substrates (i.e., arabinoxylan) is consistent with other GH62 members (32–36, 38) and continues to distinguish GH62 enzymes from arabinofuranosidases identified in other GH families.

TABLE 3.

Kinetics of SthAbf62A on pNP substrates

Substrate (μg SthAbf62A)a	Mean kcat (s−1) ± SD	Mean Km (mM) ± SD	kcat/Km (s−1 · mM−1)
pNP-α-l-arabinofuranoside (2)b	1.3 ± 0.003	3 ± 0.1	0.5
pNP-α-l-arabinopyranoside (78)b	0.01 ± 0.0002	8.2 ± 0.4	0.001

^aAmounts of purified SthAbf62A added to each reaction mixture to measure initial reaction velocities.

^bThe Michaelis-Menten equation was used for data fitting.

Regioselectivity of SthAbf62A.

¹H NMR analysis was then used to investigate the regiospecificity of SthAbf62A for α-l-Araf (1→2) and α-l-Araf (1→3) linkages in arabinoxylan. Selective removal of α-l-Araf (1→3) linked to monosubstituted β-d-Xylp by SthAbf62A was shown by the disappearance of the ¹H signal at 5.42 ppm (Fig. 2B, red line) (49, 54, 57). To clarify whether SthAbf62A activity was restricted to single α-1,3 linkages or singly substituted xylose, arabinoxylan with only single α-1,3 and α-1,2 arabinose substitutions was generated by using arabinofuranosidase (AXH-d3) from Bifidobacterium sp., which selectively removes α-l-Araf (1→3) linked to disubstituted β-d-Xylp (36, 46). The effect of AXH-d3 treatment was confirmed by the disappearance of signals near 5.28 ppm and 5.22 ppm, corresponding to the (1→3)-linked and (1→2)-linked α-l-Araf substituents of doubly substituted β-d-Xylp, and the consequent increase in the signal near 5.30 ppm, which corresponds to α-l-Araf (1→2) linked to monosubstituted β-d-Xylp (Fig. 2A) (54). Subsequent treatment with SthAbf62A significantly reduced signals at both 5.42 ppm and 5.30 ppm (Fig. 2C), consistent with the hydrolysis of both of α-1,3 and α-1,2 l-arabinofuranosyl linkages to singly substituted xylose (i.e., AXH-m2,3 activity). This analysis shows that SthAbf62A alone selectively removes singly substituted α-1,3 arabinose from wheat arabinoxylan but, similarly to PchAraf62 (36) and GH43 AXH-m2,3 from Bacillus subtilis (30), can also target singly substituted α-1,2-linked arabinose after enzymatic removal of α-1,3 l-arabinofuranosyl substituents from disubstituted xylopyranosyl residues.

FIG 2.

¹H NMR analysis of enzyme-treated wheat arabinoxylan. m1,2 and m1,3-α-l-Ara refer to monosubstituted arabinose linked to β-d-Xylp at carbon positions C-2 and C-3, respectively; d1,2 and d1,3-α-l-Ara refer to arabinose attached to double-substituted β-d-Xylp. (A) Wheat arabinoxylan (high viscosity) before and after treatment with AXH-d3 from Bifidobacterium sp. for 48 h. The effect of AXH-d3 treatment was confirmed by the disappearance of the signals near 5.28 ppm and 5.22 ppm and the concomitant increase of the signal near 5.30 ppm (54). (B) Wheat arabinoxylan (high viscosity) before and after treatment with SthAbf62A (3.0 μg) in 100 mM HEPES buffer (pH 7.0) at 45°C for 2 h. Removal of α-l-Araf (1→3) linked to monosubstituted β-d-Xylp was indicated by the disappearance of the H1 signal at 5.42 ppm; signals at 5.22 ppm and 5.28 ppm for (1→3)-linked and (1→2)-linked α-l-Araf substituents of doubly substituted β-d-Xylp remained for up to 24 h after treatment (data not shown). (C) AXH-d3-treated arabinoxylan before and after treatment with SthAbf62A (3.0 μg) in 100 mM HEPES buffer (pH 7.0) at 45°C for 2 h. Removal of (1→3)-linked and (1→2)-linked α-l-Araf substituents of singly substituted β-d-Xylp was indicated by the disappearance of the H1 signal at both 5.42 ppm and 5.30 ppm. The methylene protons at position 2 of MOPS (2.035 ppm) were used as the internal reference.

Overall structure.

Three structures of SthAbf62A were determined by molecular replacement (Table 4), which included one enzyme structure and two enzyme complex structures cocrystallized with reaction products of xylotetraose and l-arabinose, respectively (Protein Data Bank accession numbers 4O8N, 4O8O, and 4O8P). The SthAbf62A enzyme adopts a β-propeller fold with five blades (marked blades I to V) twisted and radially arranged around the pseudo-5-fold axis. The blades are linked through loops with variant lengths, and each blade was made up of four antiparallel β-strands in a typical W-like topology, with the first strand being the innermost strand (Fig. 3A). The only exception is blade IV, which was formed by a typical four-antiparallel-β-strand arrangement (β14, β15, β16, and β17) plus the small strand of β19 (Fig. 3C). This additional strand closes up one end of the β-propeller together with its flanking loop region.

FIG 3.

Overall structure of SthAbf62A revealing a five-bladed β-propeller fold. (A) Top view showing the numbering of 5 blades (blades I to V), three predicted catalytic residues, a disulfide bond between Cys297 and Cys27, a molecular Velcro (boxed in gray dashed line), and a calcium ion coordinated in the central channel of the fold. (B) Side view showing the active site surrounded by loops. (C) Overall topology of SthAbf62A highlighting the molecular Velcro between β1 and β21 (orange dashed line) and the positions of three predicted catalytic residues. β-Strands are indicated by arrows, and helices are indicated by cylinders.

Structure similarity searches of the Protein Data Bank with the Dali server showed that the overall structure of SthAbf62A resembled those of enzymes in the GH43, GH32, and GH68 families, which similarly adopt a five-bladed β-propeller fold (see Table S2 in the supplemental material). The overall structure of SthAbf62A superimposed well with the catalytic domain of these enzymes (see Table S2 in the supplemental material), which is consistent with a divergent evolution of enzymes from families GH32, -43, -62, and -68 (58). Superimposition of SthAbf62A with enzymes from these families also facilitated the prediction of the two catalytic residues (Asp53 and Glu213) and of a third residue (Asp161) believed to modulate the pK_a of the catalytic general acid (see Fig. S5 in the supplemental material) (18).

While we were completing our analysis of SthAbf62A, the first structures of GH62 α-arabinofuranosidases were reported, including UmAbf62A from Ustilago maydis, PaAbf62A from Podospora anserina, and ScAraf62A from Streptomyces coelicolor (32, 33). In line with these first structures of GH62 α-arabinofuranosidases, the five-bladed β-propeller fold of SthAbf62A is stabilized through a molecular Velcro between antiparallel β-strands 1 and 21 in blade V formed through five hydrogen bonds and several hydrophobic interactions (see Fig. S6 in the supplemental material). Belonging to subfamily GH62_2, SthAbf62A shared additional similarities to UmAbf62A and ScAraf62A. For instance, SthAbf62A contains a disulfide bond between Cys27 and Cys297, which constrains the N terminus of the enzyme in the vicinity of the loop connecting strands 19 and 20 (Fig. 3). Additionally, a calcium ion in the central channel of the fold was buried right under the l-arabinose binding pocket (Fig. 3 and 4), displaying a heptacoordination with His280, Gln216, and five water molecules.

FIG 4.

Interactions between SthAbf62A and l-arabinofuranose bound in the active-site pocket. (A) l-Arabinofuranose is shown in pink, and amino acid residues are shown in gray; polar interactions are indicated by black dotted lines. A calcium ion, shown as an orange sphere, was buried under the l-arabinose binding pocket, displaying a heptacoordination with His280, Gln216, and five water molecules, indicated as red spheres. (B) Focus on a nucleophilic water molecule, which shows shared electron density with C-1 of l-arabinose, as indicated by the black box. The electron density shown is the simulated annealing F_o − F_c omit density contoured at 2.0 σ.

Since SthAbf62A shares >65% sequence identity with ScAraf62A but demonstrates a considerably higher optimal reaction temperature and temperature stability, these enzymes were further compared in terms of 18 sequence characteristics thought to confer thermostability. Relative to ScAraf62A, SthAbf62A scored high in 11 of these 18 parameters, including a higher number of predicted hydrogen-bonding interactions and salt bridges and increased core hydrophobicity (see Table S3 in the supplemental material). Although main determinants of thermostability are still difficult to predict, the growing number of GH62 enzyme structures, including those of temperature-stable members, will undoubtedly facilitate future protein engineering efforts aimed at improving applications of GH62 enzymes.

The active site of SthAbf62A.

To understand the structural basis for catalysis and substrate binding, the structure of SthAbf62A complexes with its reaction products was obtained by separate cocrystallizations with l-arabinose and xylotetraose. A bound l-arabinose molecule was observed in a narrow and deep pocket (−1 subsite) composed of relatively conserved amino acids, which was open toward the molecular surface (Fig. 3; see also Fig. S7 in the supplemental material). It was formed mainly by the five first and innermost β-strands of each blade (β2, β6, β10, β14, and β18) and lined by about 12 amino acids (within a 4-Å distance from l-arabinose), 3 of which were acidic amino acid residues, making this pocket mainly negatively charged (see Fig. S7 in the supplemental material). The two predicted catalytic residues (Asp53 and Glu213), along with the putative pK_a modulator (Asp161), were pointing into this pocket (Fig. 3A and B) and were donated by the β2-sheet in blade I, β14 in blade IV, and β10 in blade III, respectively (Fig. 3C). The Asp53Ala and Glu213Ala substitutions led to a complete loss of enzyme activity, whereas the Asp161Ala substitution retained detectable, albeit low, activity on WAX-HV (see Table S4 in the supplemental material); the lost activity could not be rescued by NaN₃. The site-directed mutagenesis results reported here support the catalytic importance of Asp53, Glu213, and Asp161 and are consistent with mutagenesis studies of ScAraf62A that similarly implicate the corresponding Asp202, Glu361, and Asp309 residues in catalysis (33).

The l-arabinofuranose molecule was modeled with an α-configuration (Fig. 4). The ligand was anchored in the l-arabinose binding pocket mainly through several polar interactions (Fig. 4), such as hydrogen bonds between sugar ring oxygen and N^ζ of Lys52 (3.4 Å) along with O^η of Tyr314, the C-2 hydroxyl of arabinose with N^δ1 of His280 (2.9 Å) as well as O^δ1 and O^δ2 of Asp161 (2.7 Å), and the C-3 hydroxyl of arabinose with O^δ1 of Asp161 (2.6 Å) and N^ζ2 of Gln121 (2.8 Å). In particular, the contacts between the C-5 hydroxyl of l-arabinofuranose and N^ζ of Lys52 (2.8 Å) as well as O^δ2 of Asp53 (2.6 Å) might be important for discriminating between furanose and pyranose forms of l-arabinose, as observed for pNP substrates (Table 3). The catalytic Glu213 side chain was proximal to the C-1 hydroxyl of l-arabinose at a distance of 2.8 Å. All the residues involved in the above-described hydrogen bonds are contributed by β-sheets and conserved across the sequences of the GH62 family (see Fig. S8 in the supplemental material). In contrast, residues for hydrophobic interactions (i.e., Trp76, Tyr314, Ile160, and Thr68) were not highly conserved, except for Tyr314. Notably, in SthAbf62A, Trp76 along with Tyr314 appeared to contact the l-arabinose ring through edge-on-edge apolar interactions (59), which were further stabilized through interactions with the side chains of Ile160 and Thr68. In comparison, stacking interactions between arabinose and Tyr58 in UmAbf62A were predicted (32). Accordingly, binding of arabinose by SthAbf62A appears to be more similar to that by ScAraf62A, which interacts with the flat surface of arabinose through hydrophobic interactions with Ile308 (33).

The structure of SthAbf62A in complex with l-arabinofuranose revealed insights into the catalytic mechanism of this enzyme. There was a water molecule closely positioned near the anomeric carbon (2.8 Å) that also formed a hydrogen bond with the expected general base, Asp53 (Fig. 4). We also observed electron density shared between this water molecule and the anomeric carbon (Fig. 4B). Finally, the distance between Asp53 and the expected general acid, Glu213, was 8 Å. These observations are all consistent with an inverting mechanism for this enzyme, with the above-described water molecule acting as the nucleophile for attack on the anomeric carbon.

The cocrystallization of SthAbf62A with xylotetraose, a reaction product, revealed the binding mode of the xylan backbone within an open cleft situated above the catalytic pocket. Four xylose binding subsites were observed along the cleft; the xylotetraose chain fit the electron density features equally well in two opposite orientations, which is similar to the model proposed previously for acetylxylan esterases (60). The nomenclature described by McKee et al. (61) was followed here to name the subsites that bound xylotetraose. In orientation 1 of xylotetraose, the reducing-end xylose at subsite +2R was situated above the space between blades IV and V, whereas the nonreducing-end xylose at subsite +3NR was situated above the space between blades I and II (Fig. 5). Orientation 1 would allow hydrolysis of α-1,3-linked arabinose since O-3 of xylose points toward the active-site pocket, while the reverse orientation (orientation 2) would favor the hydrolysis of α-1,2-linked arabinose. In both orientations, xylose at subsite +1 was stabilized through hydrophobic stacking interactions with Tyr314 and hydrogen bonds with Glu213 (Fig. 5; see also Table S5 in the supplemental material). The nonreducing sugar unit at subsite +3NR in orientation 1, and the corresponding reducing-end sugar at subsite +3R in orientation 2, did not form any hydrogen bonds with the enzyme and instead was stabilized via a stacking interaction with Trp122, which is conserved in most sequences, with occasional substitution to Gln (see Fig. S8 in the supplemental material). Hydrogen-bonding interactions observed between the xylotetraose backbone and the enzyme are listed in Table S5 in the supplemental material. Notably, while the +2NR subsite in orientation 1 and orientation 2 is characterized by hydrogen bond interactions with Trp76 and Arg208, respectively, orientation 1 lacks the additional hydrogen bond observed for orientation 2 between Arg239 and the endocyclic oxygen of the nonreducing-end xylose. A cation, modeled as potassium, was bound in the arabinose binding site (−1 subsite) in the SthAbf62A-xylotetraose complex, and accordingly, this ion could interact with the C-3 hydroxyl (orientation 1) or the C-2 hydroxyl (orientation 2) of xylose at subsite +1 (Fig. 5). Taken together, the SthAbf62A substrate binding cleft and active site adopt a “T-like” architecture (see Fig. S9 in the supplemental material) reminiscent of that observed for two family GH43 side-chain-cleaving α-l-arabinofuranosidases (BsAXH-m2,3 from Bacillus subtilis [18] and HiAXH-d3 from Humicola insolens [61]), suggesting that the T-like architecture is a common feature of debranching glycoside hydrolases.

FIG 5.

Interactions between SthAbf62A and xylotetraose bound in the substrate binding cleft. Xylotetraose is shown in yellow, amino acids predicted to interact with the bound oligosaccharide are shown in gray, and a potassium ion bound in the arabinose binding site (−1 subsite) is shown as a purple sphere. Black dotted lines indicate polar interactions; all contacts are listed in Table S5 in the supplemental material. The xylotetraose is shown in orientation 1. All contacts remained in orientation 2, with an extra hydrogen bond between Arg239 and the endocyclic oxygen of the xylose unit at the +2NR subsite.

Conformation changes following ligand binding.

Besides the aromatic residues described above, the xylan binding cleft of SthAbf62A was lined mainly by nonconserved amino acids from loop regions, which was also noted previously for BsAXH-m2,3, UmAbf62A, PaAbf62A, and ScAraf62A (18, 32, 33), suggesting that enzymes from this clan adopt a dynamic substrate binding cleft. Consistent with this prediction, xylotetraose and l-arabinose binding induced small local conformation changes in SthAbf62A (Fig. 6). Superimpositions between the SthAbf62A structure and enzyme complexes showed pairwise root mean square deviation (RMSD) values of 0.362 Å (303 C_α atoms) and 0.173 Å (302 C_α atoms) for xylotetraose and l-arabinose complex structures, respectively.

FIG 6.

Main conformational changes of SthAbf62A upon ligand binding. Structural elements of SthAbf62A are shown in orange, the corresponding regions of the enzyme-xylotetraose complex are shown in gray, and enzyme–l-arabinose complexes are shown in green. Xylotetraose is shown in yellow, and the enzyme surface corresponding to the enzyme-xylotetraose complex is shown in gray. Superimposition of the three structures was conducted by using DaliLite.

In the SthAbf62A-xylotetraose complex structure, the difference was attributed largely to conformational differences between the five loop regions (at positions 72 to 75, 121 to 126, 148 to 160, 235 to 238, and 308 to 320), which were all involved in forming the xylan backbone binding cleft. Indeed, the pairwise RMSD values of C_α atoms computed by using only these residues increased to 0.682 Å and 0.411 Å for xylotetraose and l-arabinose complex structures, respectively, also indicating larger conformational differences induced by xylotetraose binding. Movement within these loop regions shifted the side chains of several amino acid residues, including Trp122, Asp315, and Tyr314 along with Trp76 and Arg208 from β-sheets, all of which were involved in xylotetraose binding. Furthermore, the side chain of Arg75 from the loop at positions 72 to 75 shifted toward the ligand by 2.3 Å upon xylotetraose binding (Fig. 6). Such movement within the substrate binding cleft could explain the accommodation of arabinan substrates and noncatalytic interactions between GH62 arabinofuranosidases and various oligosaccharides, including cello-oligosaccharides, which can inhibit enzyme activity (32).

Binding of l-arabinose caused similar, albeit smaller, shifts within the above-mentioned regions. However, the conformation of the l-arabinose binding pocket barely changed (pairwise RMSD values for C_α atoms of 0.277 and 0.198 Å for xylotetraose and l-arabinose complex structures, respectively).

Conclusions.

In this study, we conducted a function-structure characterization of a thermostable, bacterial α-arabinofuranosidase of the GH62 family, namely, SthAbf62A from Streptomyces thermoviolaceus. In particular, new insights into the molecular basis for substrate selectivity among GH62 enzymes were gained by studying complexes of SthAbf62A with l-arabinose and xylotetraose. All structural features of SthAbf62A underpin the strict specificity for arabinofuranosyl groups bound to backbone moieties through an alpha-glycosidic configuration. Additionally, similarly to BsAXH-m2,3 (18), the narrowness of the arabinose binding pocket likely prevents acceptance of α-l-Araf residues that are (1→2) and (1→3) linked to disubstituted β-d-Xylp, which is also consistent with ¹H NMR analyses of SthAbf62A reaction products (Fig. 2B). The structure of the complex with xylotetraose predicts at least four subsites that bind the xylan backbone, which likely contributes to the clear preference of SthAbf62A and other GH62 arabinofuranosidases for hydrolysis of polymeric arabinoxylan. Although it is unclear how neighboring arabinose substituents may influence the orientation of arabinoxylan substrates, the potential to bind xylooligosaccharides in opposite orientations explains the activity of SthAbf62A on both l-Araf substituents that are (1→2) and (1→3) linked to monosubstituted β-d-Xylp in arabinoxylan.

In summary, the recent increase in structure and functional characterizations of GH62 arabinofuranosidases reflects the particular relevance of this enzyme family for modification and characterization of polymeric and oligomeric arabinoxylans. In addition to predicting amino acid residues important for overall enzyme stability, a compelling prospect of these recent insights is to fine-tune GH62 activities through better predictions of amino acids that govern regioselectivity and accommodation of singly versus doubly substituted substrates.