Introduction

Lignocellulosic plant biomass contains polymers of cellulose, hemicellulose, and lignin bound together to form a complex. Plant biomass is considered a renewable energy resource, and carbohydrate-hydrolyzing enzymes are useful tools for biomass degradation [1]. Hemicellulose is a heteropolymer composed of a variety of sugars and typically consists of a β-1,4-xylopyranosyl backbone partially substituted with l-arabinofuranosyl, 4-O-methyl-glucuronosyl, and acetyl side chains [2]. In the case of l-arabinofuranosyl substituents, l-arabinofuranosyl residues are linked via either α-1,2- or α-1,3-bond [3] and ferulic acid is esterified to the O5 position of some l-arabinofuranosyl residues [4].

α-l-Arabinofuranosidases (EC 3.2.1.55) hydrolyze α-l-arabinofuranosyl residues from l-arabinosides and exhibit a variety of substrate specificities [5]. Most α-l-arabinofuranosidases are classified into glycoside hydrolase (GH) families GH43, GH51, GH54, and GH62 [6]. Among these, GH62 is a relatively small family, and all of the characterized GH62 enzymes are α-l-arabinofuranosidases that hydrolyze the α-1,2- or α-1,3-bond to liberate l-arabinofuranose from the xylan backbone. These enzymes have been proposed to hydrolyze sugars through an inverting mechanism [7, 8]. In recent years, three-dimensional structures of several GH62 α-l-arabinofuranosidases have been determined, and the enzymes contain the five-bladed β-propeller fold [7–10]. Similar catalytic five-bladed β-propeller domains have been found in GH43 enzymes [11, 12], and the two families, GH43 and GH62, are categorized into clan GH-F.

A basidiomycete, Coprinopsis cinerea, is known as a model mushroom-forming organism [13]. The entire genome has been sequenced [14] and shows that C. cinerea possesses three GH62 proteins, CC1G_01577, CC1G_01578, and CC1G_15259. GH62 enzymes have been phylogenetically clustered into two subfamilies, GH62_1 and GH62_2 [7]. All the three enzymes, CC1G_01577, CC1G_01578, and CC1G_15259, belong to subfamily GH62_1 and their amino acid sequences share ~60 % identity. CC1G_01577 and CC1G_15259 have a signal peptide sequence and a carbohydrate binding-module belonging to family 1 (CBM1) in their N-termini, while no signal sequence and CBM1 are found in CC1G_01578 [15]. In the previous paper, we designated CC1G_01577 as CcAbf62A and reported the cDNA cloning and characterization of this protein [15]. The results of the study indicated that the presence of a feruloyl esterase, CcEst1, enhances the activity of CcAbf62A for arabinoxylan, but the final amounts of reducing sugar in the course of arabinoxylan degradation with or without CcEst1 are observed to be almost equal. Here, we determined the crystal structure of the catalytic domain (residues 82–397) of CcAbf62A, providing insight into the structure-function relationship.

Materials and Methods

Construction of the Expression Plasmid

The cDNA of CcAbf62A was obtained as described [15]. The DNA fragment encoding amino acid residues 82–397 was amplified by PCR using the cDNA of CcAbf62A as a template and the primers 5′-TT CAT ATG CTC CCA TCC AGC TTC AGG TGG A-3′ and 5′ -T TGC GGC CGC ACA AGC GGA GTT GGT TTG AG-3′ (the restriction sites of NdeI and NotI are underlined). The amplified fragment was then ligated into the pET-21a(+) vector (Merck Millipore, Darmstadt, Germany) for heterologous expression in Escherichia coli. The resultant recombinant protein was designed to have a His-tag (AAALEHHHHHH) at the C-terminus.

Protein Expression and Purification

E. coli strain BL21 (DE3) was transformed with the obtained plasmid. The transformant was grown in 1-L Luria-Bertani (LB) medium containing 50 μg/mL ampicillin at 37 °C until the absorbance at 600 nm (A 600) reached 0.6. Protein expression was induced with isopropyl-β-d-thiogalactopyranoside at a final concentration of 0.2 mM for 18 h at 18 °C. The cells were harvested and resuspended in 30 mL of 20 mM Tris-HCl buffer (pH 7.5), followed by sonication for 2 min on ice. After centrifugation to remove insoluble material, the supernatant was applied onto a nickel (Ni2+) nitrilotriacetic acid (Ni-NTA) agarose (QIAGEN, Hilden, Germany) column equilibrated with the same buffer. The column was washed with the same buffer, and the recombinant protein was eluted with the same buffer containing 50 mM imidazole. The protein fraction was dialyzed against 10 mM Tris-HCl buffer (pH 7.5), and the purified protein yielded a single band on sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE). The concentration of the purified protein was determined by measuring the absorbance at 280 nm, using the molar extinction coefficient (1 mg/mL = 2.50) calculated by the Expasy ProtParam server (http://web.expasy.org/protparam/).

Crystallization, Data Collection, and Structure Determination

The purified protein was concentrated to 10 mg/mL in 10 mM Tris-HCl (pH 7.5) using an Amicon Ultra-15 centrifugal unit (Merck Millipore, Darmstadt, Germany). Needle-shaped crystals were obtained at 20 °C using the hanging-drop vapor diffusion method, in which 1.0 μL protein solution was mixed with an equal volume of the crystallization reservoir solution containing 0.1 M 2-(N-morpholino)ethanesulfonic acid (MES)-NaOH (pH 6.5), 0.1 M NaBr, 3 % (w/v) polyethylene glycol 20,000 (Sigma-Aldrich, St. Louis, MO, USA), and 20 % (w/v) polyethylene glycol 3350 (Hampton Research, Aliso Viejo, CA, USA). To obtain the complex structure of lead, the crystal was transferred for 20 min to a solution containing 10 mM lead(II) acetate in the reservoir solution. The harvested crystals were cryo-protected in the reservoir solution supplemented with 20 % (v/v) glycerol, and flash-frozen in liquid nitrogen. The diffraction data were collected at the AR-NE3A and AR-NW12A beamlines at the Photon Factory (Tsukuba, Japan). All data were processed and scaled using HKL2000 [16] At the time that this study was being carried out, several crystal structures of GH62 enzymes had been reported, and thus, the structure was solved by molecular replacement using the program MOLREP [17] in the CCP4 suite [18], and a model of GH62 α-L-arabinofuranosidase from Podospora anserina, Pod_ansAbf62A, (PDB id, 4N4B) was employed as a probe model. Automated model building was performed with the program ARP/wARP [19]. The model was refined using REFMAC5 [20] in the CCP4 suite, and manual adjustment and rebuilding of the model were carried out using the program COOT [21]. Validation of the structures was performed using MolProbity [22]. Figures were prepared using PyMOL (http://www.pymol.org/). The data collection and refinement statistics are summarized in Table 1. The coordinates and structure factors have been deposited in the Protein Data Bank (PDB) under the accession codes 5B6S and 5B6T.

Table 1 Data collection and refinement statistics
Full size table

Results and Discussion

Structure Determination of CcAbf62A

The crystal structure of the catalytic domain of CcAbf62A (hereafter simply referred to as CcAbf62A) was determined at 1.70-Å resolution, and that of the catalytic domain of CcAbf62A soaked with Pb(CH3COO)2 was also determined at 1.48-Å resolution (Table 1). The two structures are almost isomorphous, and thus, the following descriptions are based on CcAbf62A in the unsoaked form, unless otherwise stated.

The crystal belongs to the space group P21, which contains two molecules, Mol-A and Mol-B, in an asymmetric unit. In the structure of CcAbf62A-Pb, one lead atom is seen in the catalytic site of Mol-A (Fig. 1a). The Ramachandran plot was calculated with the MolProbity server (Table 1). Only one residue, His346, in both Mol-A and Mol-B was identified as an outlier, and the electron density for the residue was well-defined. His346 is a conserved residue in GH62, as described later, and the corresponding residue in Scytalidium thermophilum (Scy_the) Abf62C (abbreviations of GH62 enzymes are listed in Table 3) also reportedly adopts a disallowed Ramachandran conformation [10]. The 2F o-F c electron density maps contoured at 1 σ show continuous density for residues 82–398 of both Mol-A and Mol-B, and there is no significant difference between the two molecules in the catalytic cleft. Structural analysis using the PISA server [23] indicated that there was no specific interaction to form an oligomeric structure, suggesting that CcAbf62A exists as a monomer in solution. The descriptions hereafter are based on Mol-A.

Fig. 1
figure 1

Overall structure of CcAbf62A. a Ribbon model of CcAbf62A-Pb in an asymmetric unit. Mol-A (green), Mol-B (blue), glycerol molecules (red), lead atom (orange), and calcium atoms (cyan) are shown. Two glycerol molecules found near Tyr135, Tyr150, and Tyr151 are indicated with a black arrow. b Five blades I–V (blue, green, yellow, orange, and red, respectively) comprising the β-propeller fold. β1 strand (black), SS bridge (black), and calcium atom (cyan) are indicated. c Stereo view of the superimposition of the Cα backbones of CcAbf62A (red), Scy_theAbf62C (green; PDB id, 4PVI), Pod_ansAbf62A (blue; PDB id, 4N4B), and Str_coeAbf62A (yellow; PDB id, 3WN2)

Full size image

Overall Structure of CcAbf62A

The overall structure of CcAbf62A is indicated in Fig. 1b. CcAbf62A is composed of a five-bladed (blades I–V) β-propeller as observed in other GH62 enzymes. The Cα backbone of CcAbf62A was superimposed onto those of Pod_ansAbf62A (PDB ID, 4N4B) [7], Scy_theAbf62C (PDB ID, 4PVI) [10], and Streptomyces coelicolor (Str_coe) Abf62A (PDB ID, 3WN2) [8], illustrating that the folds of these enzymes are essentially identical (Fig. 1c). The β-Strands that comprise the β-propeller blades are numbered as β1-β20 (Fig. 2, top), based on the numbering scheme for Pod_ansAbf62A [7]. CcAbf62A had a disulfide bridge of Cys363-Cys397 in blade V (Fig. 1b), like those seen in Pod_ansAbf62A and Str_coeAbf62A.

Fig. 2
figure 2

Topology of the β-propeller folds of CcAbf62A and SaAraf43A. Amino acid residue numbers are given at each end of the β-strands. Eight loops located at the entrance of the catalytic cleft of the enzymes are indicated as Loop-1 to Loop-8. Loop-1 in CcAbf62A (shown in red) is longer than that in SaAraf43A, while Loop-2, Loop-4, Loop-6, Loop-7, and Loop-8 in SaAraf43A (shown in blue) are longer than those in CcAbf62A. In CcAbf62A, β-strands composed of the β-propeller fold are indicated as β1–β20, and an additional β-strand (residues 332–336) is shown. Amino acid residues, which are present in Loop-1 to Loop-8 and also listed in Table 3, are indicated

Full size image

A structural similarity search was performed by using the DALI server [24] (Table 2). Aside from the GH62 enzymes, CcAbf62A is similar to GH43, GH32, and GH68, as described previously, and these families are categorized as the GH43_62_32_68 superfamily [25]. CcAbf62A also shows homology to GH130 [26] and GH117 [27]. All of these GH families consist of five-bladed β-propeller structures. Among the characterized GH43 enzymes, CcAbf62A most resembles the Streptomyces avermitilis exo-1,5-α-l-arabinofuranosidase (SaAraf43A; PDB ID, 3AKH) [28] and the Cellvibrio japonicus 1,2-α-l-arabinofuranosidase (PDB ID, 3QEF) [29].

Table 2 Summary of structural similarity search using the DALI server
Full size table

GH62 enzymes have been proposed to be inverting glycoside hydrolases, and two Asp and one Glu residues form a catalytic triad [7, 30]. Asp109, Asp224, and Glu276 in CcAbf62A were identified as a general base, a pK a modulator, and a general acid, respectively (Table 3). With one exception [10], a calcium atom is located in the center of the five-bladed β-propeller in GH62 enzymes, and a conserved His residue holds the calcium atom (Fig. 3a). The corresponding calcium atom is also seen in CcAbf62A, and His346 functions as the calcium holder. This His residue has been proposed to form the catalytic core together with the catalytic triad residues [10].

Table 3 Comparison of amino acid residues in the substrate binding site of GH62 enzymes
Full size table
Fig. 3
figure 3

Glycerol (labeled as Gol)-bound structure of CcAbf62A. a A glycerol molecule (red) found in the active site in CcAbf62A-Pb. A calcium atom (cyan), a lead atom (orange), catalytic residues (blue), and residues interacting with calcium (green) are indicated. The electron density (F o-F c, 3σ) is shown in cyan. b Two glycerol molecules (red) found near Tyr135, Tyr150, and Tyr151. Hydrogen bonds (blue dotted line) and water molecules (magenta) are indicated. The electron density (F o-F c, 3σ) is shown in cyan. c Schematic drawing of the amino acid residues interacting with glycerol in the active site. White circle, oxygen atom; black circle, carbon atom; gray circle, nitrogen atom; dashed line, hydrogen bond

Full size image

It has been reported that Scy_theAbf62C possesses no calcium atom in the catalytic cleft, and the presence of a cysteine residue, Cys233, could result in the weak binding of the metal ion [10]. In most of the GH62 enzymes, Gln or Cys is observed in this position, whereas Asn279 is identified as the equivalent residue in CcAbf62A (Table 3). The effect of divalent cations on the activities of two enzymes from Scy. thermophilum, Scy_theAbf62A and Scy_theAbf62C, has been investigated. The presence of Ca2+ or Mg2+ resulted in only small changes in the activities of the two enzymes, while the presence of 2 mM of Ni2+, Co2+, Zn2+, Cu2+, or Mn2+ inhibited both enzymes. In particular, a significant decrease was observed in the presence of Zn2+ or Cu2+, both of which have a relatively large atomic radius [10]. The structure of CcAbf62A-Pb Mol-A shows that Pb2+ binds to the catalytic acid residue, Glu276, and does not interact with the calcium holder, His346 (Fig. 3a). It is likely that metal atoms having relatively large radii occupy positions different from that of the calcium atom.

Seven and six glycerol molecules from the cryoprotectant solution were identified in the CcAbf62A and CcAbf62A-Pb asymmetric units, respectively. A glycerol molecule each is located in the catalytic cleft of both Mol-A and Mol-B (Fig. 1a) and forms the same contacts with CcAbf62A (Fig. 3a). The other glycerol molecules were found on the protein surface (Fig. 1a). It is likely that some of these glycerol molecules are artifacts, as they are present near the crystal contacts. Recently, however, a secondary carbohydrate binding site, composed of Trp23 and Tyr44, has been identified in Aspergillus nidulans α-l-arabinofuranosidase, a member of the GH62_2 subfamily [31]. The two residues are not conserved in CcAbf62A; instead, Tyr135, Tyr150, and Tyr151 of CcAbf62A are located near the corresponding position of the secondary carbohydrate binding site in A. nidulans α-l-arabinofuranosidase, and two glycerol molecules in Mol-A are located close to these Tyr residues (Fig. 3b). The three Tyr residues are also found in the GH62_1 subfamily enzymes, Pod_ansAbf62A and Scy_theAbf62C, suggesting that the three Tyr residues might participate in substrate binding.

Ligand-Bound Model of CcAbf62A

Several ligand-bound structures of GH62 enzymes have been reported. We constructed a ligand-bound model of CcAbf62A using Ustilago maydis (Ust_may) Abf62A complexed with α-l-arabinofuranose (Ara) (PDB ID, 4N2R) [7], and Str_coeAbf62A complexed with xylopentose (X5) [8] (PDB ID, 3WN2). The three structures were superimposed and the coordinates of Ara and X5 were then placed in CcAbf62A. To probe the amino acid residues involved in each subsite, residues within 4 Å from Ara or X5 were calculated using the program NCONT of CCP4, and the corresponding residues in other GH62 enzymes are listed (Table 3). The subsite numbers are based on those described [8]; Ara is accommodated by subsite −1, the enzymatic cleavage occurs between subsites −1 and +1, and X5 interacts with subsites +4NR, +3NR, +2NR, +1, and +2R from the non-reducing end to the reducing end. Amino acid residues potentially located at subsite +3R and those potentially interacting with the calcium atom, namely the residues corresponding to Arg211 in Pod_ansAbf62A and those corresponding to Cys233 in Scy_theAbf62C, are also listed in Table 3.

A glycerol molecule binds at subsite −1 in both Mol-A and Mol-B of CcAbf62A (Fig. 3a), and the structure O1-C1-C2-(O2)-C3-O3 of glycerol is similar to that formed by atoms O3-C3-C4-(O4)-C5-O5 of Ara (Fig. 4a, b). The interaction between CcAbf62A and glycerol was analyzed with the programs Ligplot [32] and Coot, indicating that as many as 11 residues participate in the binding of glycerol (Fig. 3c). Based on the ligand-bound model of CcAbf62A, the amino acid residues forming hydrogen bonds with glycerol (Lys108, Asp109, Gln181, Asp224, Glu276, Glu294, His346, Gln370, and Tyr380) appear to participate in subsite −1, and these residues are fully conserved among GH62 enzymes (Table 3).

Fig. 4
figure 4

Substrate bound model of CcAbf62A. a Cα backbone representation of CcAbf62A. Models of α-l-arabinofuranose (yellow) and xylopentaose (magenta) are placed on the structure. Red stick, glycerol; blue stick, catalytic residue; cyan ball, calcium atom. b, c Comparison of some key residues interacting with the xylan backbone in CcAbf62A (b) and those in Str_theAbf62A (PDB id, 4O8P) (c). Magenta stick, xylopentaose model (b) or xylotetraose (c); yellow stick, α-L-arabinofuranose model; red stick, glycerol. Subsite numbers (−1, +1, +2R, +2NR, +3NR, and +4NR) are indicated

Full size image

In the previous research, a feruloyl esterase, CcEst1, was found to promote the activity of CcAbf62A against feruloyl arabinoxylan, whereas the amount of reducing sugar in the late stage of the reaction was the same regardless of the presence or absence of CcEst1 [15]. There is a space, which could potentially accommodate a ferulate residue, around atoms C5–O5 of Ara in the ligand-bound model of CcAbf62A, as two amino acid residues surrounding the space are identified as tyrosine residues, Tyr131 and Tyr161 (Fig. 4b), whereas bulky Trp residues are found in the corresponding positions in some of the enzymes (Table 3). For example, the corresponding two residues in Streptomyces thermoviolaceus (Str_the) Abf62A are identified as Trp111 and Trp157 (Fig. 4c).

While the residues in subsites −1 and +1 are well conserved among GH62 enzymes, relatively large variations are found in subsites +3R, +2NR, +3NR, and +4NR. It is interesting to note that a His residue, His221, is not conserved in the other GH62 enzymes, and the corresponding residue is mostly Tyr or Thr. The imidazole ring of His221 stacks against the pyranose ring of xylose at subsite +2NR, and His221 appears to be the most critical residue for the binding of subsite +2NR (Fig. 4b). In contrast, the residue at the equivalent position in Str_theAbf62A, Thr192, does not directly interact with the xylan backbone (Fig. 4c). We have reported that CcAbf62A does not hydrolyze p-nitrophenyl α-l-arabinofuranoside [15], which is commonly used for measurement of α-l-arabinofuranosidase activity. Site-directed mutagenesis of Scy_theAbf62C indicated that alteration of Tyr168, the residue corresponding to His221 in CcAbf62A, leads to decrease of the activity for p-nitrophenyl α-l-arabinofuranoside [10]. Also, comparative properties of two GH62_1 α-l-arabinofuranosidases, ABFI and ABFII, from Aspergillus fumigatus have been reported [33]; ABFI has a very high K m value, 94 mM for p-nitrophenyl α-l-arabinofuranoside, whereas ABFII exhibits a low K m value, 3.9 mM, for the same substrate. The residues equivalent to His221 in CcAbf62A are identified as Asn157 (ABFI) and Tyr222 (ABFII), suggesting that the presence of Tyr residue may be critical for the activity for p-nitrophenyl α-l-arabinofuranoside. It is likely that the presence of the His residue in subsite +2NR may result in the inactivity of CcAbf62A for the substrate p-nitrophenyl α-l-arabinofuranoside.

Sequence alignment of the three GH62 enzymes from C. cinerea: CcAbf62A, CC1G_01578, and CC1G_15259, indicates that CC1G_01578 and CC1G_15259 have additional typical amino acid residues found in other GH62 enzymes; the corresponding residues of His221 and Asn279 in CcAbf62A are identified as Tyr (CC1G_01578, Tyr178; CC1G_15259, Tyr222) and Cys (CC1G_01578, Cys236; CC1G_15259, Cys280), respectively. It is likely that the subsite affinities of CcAbf62A for the xylan backbone, and the effect of metal ions are different from those of other GH62 enzymes, including CC1G_01578 and CC1G_15259.

Comparison to GH43 Enzymes

The similarity of the β-propeller fold between GH62 and GH43 has been documented. As CcAbf62A showed high structural similarity with SaAraf43A [28] among the GH43 enzymes in the DALI search, the structures of CcAbf62A and SaAraf43A were compared. The β-strand backbones, which consist of the five-bladed β-propeller structures of CcAbf62A and SaAraf43A, are structurally identical, while the positions of the N- and C-termini were different between the two structures. The first β-strand, designated β1, of CcAbf62A is present at the identical position to the fourth β-strand of SaAraf43A, resulting in formation of a so-called “molecular Velcro” [7] at blade V in CcAbf62A (Fig. 1b). Also, an additional short β-strand, comprising residues 332–336, forms a parallel β-sheet with β1 in blade 1 of CcAbf62A (Fig. 2, top).

Eight loops located at the entrance of the catalytic cleft of CcAbf62A adopt structures different from those of SaAraf43A, and these loops are designated Loop-1 to Loop-8 (Fig. 2). All the amino acid residues in subsites +3R, +2R, +1, +2NR, +3NR, and +4NR, which interact with the xylan backbone, are found in these eight loops. It is interesting to note that the lengths of Loop-2, Loop-4, Loop-6, Loop-7, and Loop-8 in CcAbf62A are shorter than the corresponding loops in SaAraf43A, which allows the molecular surface of CcAbf62A to form a xylan binding cleft. In contrast, Loop-1 in CcAbf62A appears to be longer than that in SaAraf43A (Fig. 5). Loop-1 appears to be critical for the enzymatic activity of CcAbf62A, since Tyr380 is located in Loop-1. Tyr380 forms hydrogen bonds with the catalytic residue, Asp109, via a water molecule (Fig. 3c), and this Tyr residue is strictly conserved among GH62 enzymes (Table 1). Mutation of the corresponding residue in Str_coeAbf62A, Tyr461, has been reported to cause a drastic decrease in activity [8].

Fig. 5
figure 5

Stereo view of the superimposition of the Cα backbones of CcAbf62A (pink) and SaAraf43A (cyan). Loop-1 in CcAbf62A (shown in red) is longer than that in SaAraf43A, whereas Loop-2, Loop-4, Loop-6, Loop-7, and Loop-8 in SaAraf43A (shown in blue) are longer than those in CcAbf62A

Full size image

Conclusions

The crystal structure of CcAbf62A was determined in this study. The structure reveals that residues in subsites +3R, +2NR, +3NR, and +4NR of CcAbf62A are relatively not conserved compared to those of other GH62 enzymes. In particular, a His residue, His221, is uniquely found in subsite +2NR of CcAbf62A, which may be responsible for the inactivity for p-nitrophenyl α-l-arabinofuranoside. In addition, although a calcium atom is observed in the structure of CcAbf62A, the residue interacting with calcium is identified as an Asn residue, Asn279, which is different from the Cys or Gln residues found in other GH62 enzymes. Because of the lack of the bulky Trp residue in the substrate binding site, there is a space near the catalytic center of CcAbf62A, which is likely to be capable of accommodating feruloyl l-arabinose.

There are three GH62 enzymes, CcAbf62A, CC1G_01578, and CC1G_15259 from C. cinerea, and two of them, CcAbf62A and CC1G_15259, possess a carbohydrate-binding module belonging to CBM1 in their N terminus. In the majority of the GH62 enzymes, however, no CBM1 module was found, as shown in Table 3, and CC1G_01578 is more like a typical α-l-arabinofuranosidase. CcAbf62A therefore appears to be structurally unique, despite its high structural similarity to other GH62 enzymes. It is not uncommon for a fungal genome to possess multiple GH62 genes [10, 33]. The results obtained here suggest that amino acid residues interacting with the xylan backbone are not conserved among the GH62 enzymes; this strategy is likely to be suitable for hydrolyzing a wide variety of arabinoxylan structures by multiple α-l-arabinofuranosidases.