Structural and functional characterization of a bifunctional GH30-7 xylanase B from the filamentous fungus Talaromyces cellulolyticus

Glucuronoxylanases are endo-xylanases and members of the glycoside hydrolase family 30 subfamilies 7 (GH30-7) and 8 (GH30-8). Unlike for the well-studied GH30-8 enzymes, the structural and functional characteristics of GH30-7 enzymes remain poorly understood. Here, we report the catalytic properties and three-dimensional structure of GH30-7 xylanase B (Xyn30B) identified from the cellulolytic fungus Talaromyces cellulolyticus. Xyn30B efficiently degraded glucuronoxylan to acidic xylooligosaccharides (XOSs), including an α-1,2-linked 4-O-methyl-d-glucuronosyl substituent (MeGlcA). Rapid analysis with negative-mode electrospray-ionization multistage MS (ESI(−)-MSn) revealed that the structures of the acidic XOS products are the same as those of the hydrolysates (MeGlcA2Xyln, n > 2) obtained with typical glucuronoxylanases. Acidic XOS products were further degraded by Xyn30B, releasing first xylobiose and then xylotetraose and xylohexaose as transglycosylation products. This hydrolase reaction was unique to Xyn30B, and the substrate was cleaved at the xylobiose unit from its nonreducing end, indicating that Xyn30B is a bifunctional enzyme possessing both endo-glucuronoxylanase and exo-xylobiohydrolase activities. The crystal structure of Xyn30B was determined as the first structure of a GH30-7 xylanase at 2.25 Å resolution, revealing that Xyn30B is composed of a pseudo-(α/β)8-catalytic domain, lacking an α6 helix, and a small β-rich domain. This structure and site-directed mutagenesis clarified that Arg46, conserved in GH30-7 glucuronoxylanases, is a critical residue for MeGlcA appendage–dependent xylan degradation. The structural comparison between Xyn30B and the GH30-8 enzymes suggests that Asn93 in the β2–α2 loop is involved in xylobiohydrolase activity. In summary, our findings indicate that Xyn30B is a bifunctional endo- and exo-xylanase.

In contrast to bacterial enzymes, there are very few reports on fungal GH30-7 xylanases. Cellulolytic fungi, such as Trichoderma reesei, Myceliophthora thermophila, and Talaromyces cellulolyticus, which are promising enzyme sources for hydrolyzing lignocellulosic biomass (14 -16), encode multiple putative GH30-7 xylanases in their genomes. The GH30-7 xylanases are secreted in cellulosic and xylanosic culture conditions (17,18). However, information on their catalytic properties is limited, except for those expressed in T. reesei. This fungus produces two types of GH30-7 xylanases possessing exo-xylanase activity toward the reducing end of xylan (XYN IV) and glucuronoxylanase activity similar to the bacterial GH30-8 enzyme (XYN VI) (19,20). The notable difference in the sequences of fungal GH30-7 and GH30-8 xylanases is the absence of an Arg 293 counterpart (20). Without a three-dimensional structure of GH30-7, it is hard to determine the structural underpinnings of the MeGlcA appendage dependence of XYN VI and structural differences between exoxylanase and glucuronoxylanase.
In a preliminary study, we detected two GH30-7 proteins from T. cellulolyticus, termed Xyn30A (NCBI protein ID GAM43270) and Xyn30B (GAM36763), which were secreted as major proteins in a culture containing birchwood glucuronoxylan (Fig. S1). Xyn30A has been predicted to be a putative exoxylanase from its relatively high sequence similarity with XYN IV (77%), whereas Xyn30B has remained to be identified due to lack of an appropriate GH30-homologous protein. Here, we report the catalytic properties and crystal structure of Xyn30B. Xyn30B exhibited an obvious, MeGlcA appendage-dependent glucuronoxylanase activity. Moreover, we found that Xyn30B exhibits a novel xylobiohydrolase activity wherein xylobiose (Xyl 2 ) is released from the nonreducing end of ␤-1,4-xylan and XOS. This unique bifunctional activity and the conserved Arg residue found in Xyn30B are discussed based on the structural comparison between GH30-8 and GH30-7 xylanase.

Amino acid sequence analysis of Xyn30B
The Xyn30B gene is composed of 1,425 bp without introns in the T. cellulolyticus genome and encodes a protein consisting of 474 amino acid residues. Xyn30B has a relatively high amino acid sequence identity with fungal GH30-7 enzymes, such as XYN IV (38.2%) and XYN VI (42.2%) from T. reesei and XYLD (53.4%) from Bispora sp. (19 -21), as compared with bacterial GH30-8 enzymes, such as EcXynA (24.4%), BsXynC (23.3%), and CaXyn30A from Clostridium acetobutylicum (26.3%) (2,12,22). Two conserved catalytic residues previously identified in GH30 xylanase-a general acid/base residue and a nucleophilic residue-were found to correspond with Glu 202 and Glu 297 , respectively, in Xyn30B (Fig. 1, gray highlights). As with XYN IV and XYN VI, an Arg residue responsible for recognition of the MeGlcA in GH30-8 enzymes is not conserved in Xyn30B (Fig. 1, red box). The Xyn30B amino acid sequence includes a signal sequence (residues 1-22) as predicted by the SignalP server (23). The cleavage site of the signal peptide was estimated to lie between Ala 19 and Ile 20 or between Ala 22 and Gln 23 . Eight of the N-glycosylation sites (Asn 60 , Asn 88 , Asn 111 , Asn 154 , Asn 215 , Asn 334 , Asn 346 , and Asn 412 ) were predicted by the NetNglyc server (http://www.cbs.dtu.dk/services/NetNGlyc/). 3 Xyn30B was overexpressed using the T. cellulolyticus homologous expression system (24). SDS-PAGE analysis of the purified enzyme showed a molecular mass slightly larger than 49,403 Da, which has been estimated from the primary structure excluding the N-terminal signal peptide (Fig. 2). Furthermore, the average molecular mass of Xyn30B determined by TOF-MS was 56,354 Da, indicating that Xyn30B was glycosylated at several sites. The glycosylation sites in Xyn30B were assigned by X-ray crystallography, as described below.

Enzyme characterization
Xyn30B exhibited xylanase activity on beechwood xylan (11.3 units mg Ϫ1 ) and birchwood xylan (9.0 units mg Ϫ1 ), whereas degradation activities for arabinoxylan, carboxymethyl cellulose, glucomannan, and xyloglucan were not detected by the 3,5-dinitrosalicylic acid (DNS) method. The optimum pH and temperature for hydrolysis of beechwood xylan were estimated around pH 4 and 50°C, respectively (Fig. S2). Xyn30B retained more than 90% activity after incubation for 30 min at 40°C in pH over a range of 3-6.5 and was stable for 24 h at temperatures below 40°C at pH 4.0.
The initial degradation product of beechwood xylan by Xyn30B was found to be acidic XOSs, whereas linear oligosaccharides and xylose were hardly detected (Fig. 3). These observations indicate that Xyn30B is a glucuronoxylan-specific xylanase. Xyn30B also degraded the MeGlcA-appended oligosaccharide analogue, borohydride-reduced aldotetrauronic acid (BR-MeGlcA 3 Xyl 3 ), into aldotriuronic acid (MeGlcA 2 Xyl 2 ) and xylitol. Kinetic parameters were also determined as follows: K m ϭ 19 mg ml Ϫ1 and k cat ϭ 17 s Ϫ1 for beechwood xylan; K m ϭ 0.064 mM and k cat ϭ 23 s Ϫ1 for BR-MeGlcA 3 Xyl 3 . The low K m value for BR-MeGlcA 3 Xyl 3 suggests that Xyn30B has high affinity for MeGlcA.

Molecular and structural characterization of the acidic XOS products
The molecular content in acidic XOS produced by Xyn30B was evaluated by ESI(Ϫ)-MS (Fig. 4A). Acidic products were readily observed as singly and doubly charged deprotonated species generically labeled [Xyl n MeGlcA Ϫ H] Ϫ (1Ϫ, mainly COO Ϫ from MeGlcA) and [Xyl n MeGlcA Ϫ 2H] 2Ϫ (2Ϫ, COO Ϫ from MeGlcA and O Ϫ from anomeric carbon). They correspond to the XOS backbones formed by n xylose units and carrying one MeGlcA moiety with no information about its position. The ESI(Ϫ) analysis filters the oligoxylose species, Xyl n , carrying no acidic moiety and allows for instant visualization of the shortest acidic product, which was found to be Xyl 2 MeGlcA at m/z 471 (Fig. 4A, highlighted in boldface red) and associated with a broad distribution of longer congeners up to Xyl 14 MeGlcA at m/z 1,027 (2Ϫ).
The position of the MeGlcA moiety along the XOS chain (reducing end, nonreducing end, or in between) was further revealed using a multistage MS procedure (MS n ). Upon activa-tion in MS 2 (Fig. 4B), the [Xyl 2 MeGlcA Ϫ H] Ϫ expelled a neutral C 2 H 4 O 2 via a cross-ring cleavage, yielding 0,2 A 3 at m/z 411 (Ϫ60 Da) concomitantly to a xylose unit via a glycosidic bond The features are shown as follows: conserved Arg residues in GH30-8 glucuronoxylanases (red box); catalytic Glu residues (highlighted in gray); predicted signal sequence of Xyn30B (highlighted in black); ␤2-␣2 loop of Xyn30B and EcXynA (boldface type); N-glycosylated Asn residues (red characters); the conserved Arg residues in GH30-7 glucuronoxylanases (highlighted in red); Asn 93 of Xyn30B (indicated by a black arrow). Secondary structures of Xyn30B are shown above the sequences. Each of the secondary structures of EcXynA and BsXynC is also indicated within the sequence. N-terminal signal sequence, N-glycosylation sites, and secondary structures are based on assignment by the crystal structure.

Characterization and crystal structure of Xyn30B
cleavage that yielded a C 2 at m/z 339 (Ϫ132 Da). As the precursor ion is two xylose units long, it instantly indicates that the MeGlcA is located at the nonreducing end of Xyl 2 (Fig. 4C). Both cross-ring and glycosidic bond cleavages have been found to occur only at the reducing ends of deprotonated acidic products (25,26). The ESI(Ϫ)-MS 3 spectrum of C 2 displays two product ions formed upon its dehydration (B 2 at m/z 321, Ϫ18 Da) and a cross-ring cleavage of the last xylose moiety ( 0,2 X 2 at m/z 249) as the two sole fragmentation pathways that are opened up by the MeGlcA position at the reducing end of the activated species (Fig. 4B).
Both the cross-ring cleavage (yielding 0,2 A 4 at m/z 543) and the glycosidic bond cleavage (yielding C 3 at m/z 471) were observed in the ESI(Ϫ)-MS 2 fingerprint of the deprotonated acidic product, [Xyl 3 MeGlcA Ϫ H] Ϫ , at m/z 603 (Fig. 5A), indicating that the MeGlcA residue is not located at its reducing end. In the MS 3 spectra (Fig. 5A), C 3 was found to dissociate into B 3 at m/z 453 (dehydration, Ϫ18 Da) and 0,2 X 2 only (crossring cleavage at the reducing end). Deviating from the MS 2 spectrum of [Xyl 2 MeGlcA Ϫ H] Ϫ (Fig. 4B) but resembling the MS 3 pattern of C 2 from [Xyl 2 MeGlcA Ϫ H] Ϫ , it unambiguously localized the acidic MeGlcA pendant group at the reducing end of C 3 . In a reverse chain reconstruction, one xylose unit was added to the reducing end of C 3 , demonstrating that [Xyl 3 MeGlcA Ϫ H] Ϫ carries the glucuronic acid moiety one unit away from the reducing end (Fig. 5C).
A similar conclusion was drawn for [Xyl 5 MeGlcA Ϫ H] Ϫ at m/z 867 (Fig. 5B); the ESI(Ϫ)-MS 2 spectrum barely displayed a xylose-shorter C 5 ion product at m/z 735 (glycosidic bond cleavage at the reducing end), indicating the MeGlcA is not located at the reducing end but probably one unit away, considering the low intensity of C 5 (26). Its ESI(Ϫ)-MS 3 fingerprint was eventually identical to the previous MS 3  This indicates a generic MeGlcA 2 Xyl n (n Ͼ 1) shape for the all of the acidic products released using Xyn30B (Fig. 4A). From these results, Xyn30B appears to specifically cleave glucuronoxylan at the second glycosidic linkage from the MeGlcA residue toward the reducing end, similarly to typical GH30 glucuronoxylanases.

Xylobiohydrolase activity
During the hydrolysis of beechwood xylan by Xyn30B, we noticed that Xyl 2 was produced after prolonged incubation and increased protein loading of the reaction mixture (Fig. 6A). The increases in MeGlcA 2 Xyl 2 and MeGlcA 2 Xyl 3 were also observed with a decrease in longer acidic XOS in the mixture. These results indicate that Xyl 2 was produced by further degradation of the acidic XOS. The specific production of Xyl 2 suggests that Xyn30B has xylobiohydrolase activity, releasing Xyl 2 units from the acidic XOSs that were generated in the initial stage of the reaction. This activity was predicted to release the product from the nonreducing end of MeGlcA 2 Xyl n . In addition, the production of xylotetraose (Xyl 4 ) and xylohexaose (Xyl 6 ) was significantly increased with increasing time compared with that of xylotriose (Xyl 3 ) and xylopentaose (Xyl 5 ) (Fig. 6A). The product concentrations are shown in Table S1. These observations imply that the xylobiohydrolase activity was accompanied by transglycosylation activity, which transfers the Xyl 2 from the acidic XOS to the free acceptors (Xyl 2 and Xyl 4 ).
The xylobiohydrolase activity of Xyn30B was also confirmed for linear XOS, which was MeGlcA appendage-independent. When Xyl 3 was used as the substrate, xylose and Xyl 2 were produced as hydrolysates, and Xyl 5 was formed through a transglycosylation (Fig. 6B). The hydrolase and transglycosylation activities of Xyl 3 were 0.388 and 0.303 units mg Ϫ1 for the production of xylose and Xyl 5 , respectively. The major products from Xyl 4 were identified as Xyl 2 and Xyl 6 . A small amount of unidentified XOS longer than Xyl 6 was also observed during the hydrolysis of these substrates, probably due to further transglycosylation (Fig. 6B, arrows). In contrast, no products were pro-  The reaction mixture containing 10 mg ml Ϫ1 beechwood xylan, 50 mM sodium acetate, pH 4.0, and 10 g ml Ϫ1 Xyn30B was incubated at 40°C for 1 h followed by incubation at 99°C for 5 min to stop enzyme reaction.

Characterization and crystal structure of Xyn30B
duced when only Xyl 2 was used as the substrate (data not shown), suggesting that transglycosylation occurs during the hydrolysis. These results indicate that Xyn30B is a bifunctional enzyme possessing both MeGlcA appendage-dependent glucuronoxylanase activity and xylobiohydrolase (including transglycosylation) activity.

X-ray crystallography of Xyn30B
The crystal structure of Xyn30B was determined at 2.25 Å resolution by molecular replacement using CaXyn30A as the search model (PDB code 5CXP). The data collection and refinement statistics are shown in Table 1. The Xyn30B crystal was in the P2 1 2 1 2 1 space group with two protein molecules (chains A and B) in the asymmetric unit. Amino acid residues numbered 20 -473 were assigned to chains A and B with the electron density map indicating that the N-terminal signal sequence was cleaved between Ala 19 and Ile 20 . Glu 474 , which is the C-terminal residue, could not be assigned due to disorder.
To clarify the Xyn30B substrate recognition mechanism, the crystal structure was superimposed on the EcXynA model complexed with MeGlcA 2 Xyl 3 and imidazole (PDB code 2Y24) (Fig.  8A). In the EcXynA-ligand complex, three xylose residues of MeGlcA 2 Xyl 3 are located in the subsite Ϫ1, Ϫ2a, and Ϫ3; a MeGlcA residue is in the subsite Ϫ2b; and an imidazole is in the putative ϩ1 subsite (Fig. 8B) (9). Amino acid residues in the ϩ1 and Ϫ1 subsites of the two enzymes are highly conserved except for a few variations (Fig. 8, C and D). The residues corresponding to Trp 141 , Asn 201 , Glu 202 , Tyr 209 , Tyr 279 , Glu 297 , and Trp 341 of Xyn30B are found in the subsites of EcXynA. Trp 168 and Leu 204 of EcXynA, which probably interact with substrates at the ϩ1 subsite, are substituted by Glu 205 and Val 245 in Xyn30B, respectively (Fig. 8C).
The notable differences between Xyn30B and EcXynA are found at subsites Ϫ2b, Ϫ2a, and Ϫ3. Subsite Ϫ2b of Xyn30B is formed by Arg 46 , Leu 301 , Trp 341 , Ile 342 , Glu 345 , Thr 349 , and Ser 351 (Fig. 8E). The residues are substantially different between Xyn30B and EcXynA except for Trp 341 , corresponding to EcXynA Trp 289 (Fig. 8, E and F). Interestingly, the guanidinium group of Arg 46 in Xyn30B is located in the same region as Arg 293 in EcXynA (Fig. 8, E and F). EcXynA Arg 293 is the important residue that forms an ionic interaction with the carboxyl group of the MeGlcA side chain (9 -11). Therefore, Xyn30B Arg 46 has been suggested to have the same role in position Ϫ2b as Arg 293 found in bacterial GH30-8 enzymes.
Subsites at the Ϫ3 position of Xyn30B have limited space with the presence of the ␤2-␣2 loop composed of Gly 85 -Tyr 117 (Fig. 8, G and I) in contrast to that of EcXynA (Fig. 8, H and J). Binding of the nonreducing end of MeGlcA 2 Xyl 3 seems to be

Characterization and crystal structure of Xyn30B
blocked at the Ϫ3 position by the ␤2-␣2 loop (Fig. 8I). This loop is significantly longer than the loop of EcXynA ( Fig. 1 and  Fig. S5). Asn 93 on the tip of the loop is more specific in Xyn30B and has been predicted to be located proximal to the xylose at the Ϫ2a subsite (Fig. 8G). These observations suggest that the substrate binding at the Ϫ2a and Ϫ3 positions in Xyn30B is apparently different from that in EcXynA under the influence of the ␤2-␣2 loop.

Site-directed mutagenesis
The role of Arg 46 in Xyn30B was confirmed by site-directed mutagenesis by substituting Arg 46 for Ala (Xyn30B R46A). The specific activity of R46A for beechwood xylan was 3.52 units mg Ϫ1 , which was ϳ3.2-fold lower than that of the WT enzyme. The K m value of R46A for BR-MeGlcA 3 Xyl 3 was not determined in this study, because the initial rate of xylitol production was not saturated, even at the substrate concentration of 24 mM. These results suggest that Arg 46 within Xyn30B plays an important role in the recognition of MeGlcA side chain.
In contrast, the xylobiohydrolase and transglycosylation activities for Xyl 3 were 0.505 units mg Ϫ1 and 0.371 units mg Ϫ1 , respectively, which were both slightly higher than the WT enzyme. The degradation pattern of beechwood xylan by R46A exhibited significantly high production of xylobiose (Fig. 6C). Considering the hydrolytic activity of Xyn30B and R46A for Xyl 3 , acidic XOS appears to act as an inhibitor as well as a substrate during xylobiohydrolysis in the WT enzyme.
The endo-xylanase activity of R46A was suggested to be glucuronoxylan-specific, similar to the WT enzyme, based on the HPAEC elution pattern of acidic XOS in a reaction mixture wherein enzyme loading was increased (data not shown). The endo-xylanase activity for arabinoxylan was also not detected for R46A. These observations agree with previous reports suggesting that mutants of the conserved Arg residue in the bacterial GH30-8 enzymes, EcXynA and StXyn30A, which is the GH30 glucuronoxylanase from Streptomyces turgidiscabies, degrade glucuronoxylan in the same manner as the WT enzymes, whereas their catalytic efficiencies were lower (8, 11).

Characterization and crystal structure of Xyn30B
Discussion This paper described the characterization and structural determination of the novel GH30-7 xylanase, Xyn30B, from T. cellulolyticus. Xyn30B displayed glucuronoxylan-specific endo-xylanase activity that has been reported in bacterial GH30-8 glucuronoxylanases and T. reesei GH30-7 XYN VI. X-ray crystallography of Xyn30B and site-directed mutagenesis revealed that Arg 46 is important for recognition of the MeGlcA residue, corresponding to a conserved Arg 293 of GH30-8 EcXynA. Xyn30B was demonstrated to release Xyl 2 units from the nonreducing end of the acidic XOS and linear XOS in an exo-fashion and also to transfer Xyl 2 to an aglycon receptor (Fig. 6, A and B). To our knowledge, this is the first evidence describing exo-type xylobiohydrolase activity. Although xylobiosyltransferase activity has been reported for GH10 endoxylanase (28 -30), xylobiose-specific transglycosylation in Xyn30B is expected to be more profitable for producing xylobioside products. A study of the transglycosylation reaction of Xyn30B is in progress. A dual function showing endo-/exoxylanase activity has not been reported for bacterial GH30-8 glucuronoxylanase, whereas XYN VI is capable of slower but significant cleavage of unsubstituted parts of xylan and acidic XOS (20). The catalytic diversity observed in Xyn30B and XYN VI may be a common property of fungal GH30-7 glucuronoxylanases.
To confirm the generality of Arg 46 in GH30-7 xylanases, the amino acid sequences of GH30-7 enzymes assigned in the CAZy database were aligned with Xyn30A and Xyn30B from T. cellulolyticus using molecular phylogenetic analysis. As of December 2018, 20 fungal enzymes and 22 actinobacterial enzymes are assigned to the GH30-7 subfamily in the CAZy database. Three enzymes were excluded for the analysis due to their lack of catalytic Glu residues. The sequence alignment revealed that GH30-7 and GH30-8 enzymes are obviously distributed into different clusters in the molecular phylogenetic tree and that GH30-7 enzymes are further divided into fungal and actinobacterial groups (Fig. 9). We found that Arg 46 in Figure 6. HPAEC-PAD profiles of Xyn30B reaction products. A, time course analysis of Xyn30B products from beechwood xylan. Hydrolysis was performed at 40°C in a mixture consisting of 10 mg ml Ϫ1 beechwood xylan and 100 g ml Ϫ1 Xyn30B in 50 mM sodium acetate (pH 4.0). B, XOSs from Xyl 3 (top) and Xyl 4 (bottom) produced by Xyn30B. The hydrolysis of 10 mM Xyl 3 and 10 mM Xyl 4 was performed using 100 g ml Ϫ1 Xyn30B in 50 mM sodium acetate (pH 4.0) at 40°C for 60 min. Arrows indicate linear XOSs that are longer than Xyl 6 . C, comparison of the degradation profiles of beechwood xylan by Xyn30B and R46A. Hydrolysis was performed at 40°C for 60 min in a mixture consisting of 10 mg ml Ϫ1 beechwood xylan and 20 g ml Ϫ1 enzyme in 50 mM sodium acetate (pH 4.0).

Characterization and crystal structure of Xyn30B
Xyn30B is a highly conserved residue in the greater part of fungal and actinobacterial GH30-7 enzymes (Fig. S6). Fungal enzymes, which have a residue corresponding to Arg 46 of Xyn30B, form a large cluster that includes Xyn30B and XYN VI ( Fig. 9; Arg 46 is conserved), suggesting that these enzymes share relatively high amino acid sequence similarity and therefore are glucuronoxylanases. This group included only one exception, Fusarium fujikuroi CCT73001, which has a His residue instead of an Arg residue. However, the His residue may form an ionic interaction with the MeGlcA residue of glucuronoxylan in the same manner as the Arg residue, because the side chain of His is positively charged in acidic conditions, which is a common condition of fungal xylanase. In contrast, XYN IV, Xyn30A, Thielavia terrestris THITE_2123443, and Actinoplanes derwentensis SDT08346, which do not possess a residue corresponding to Arg 46 of Xyn30B, were located in an independent cluster ( Fig. 9; Arg 46 is not conserved). It seems reasonable that XYN IV does not have the Arg residue, because XYN IV is known to be an exo-xylanase but not a glucuronoxylanase (19). Xyn30A, T. terrestris THITE_2123443, and A. derwentensis SDT08346 are predicted to have enzyme activities that differ from glucuronoxylanase. Unique structural features in the crystal structure of Xyn30B, including a ␤2-␣2 long loop, a ␤-sheet structure composed of ␤8-, ␤8A-, and ␤8B-strands (Fig. 7), and an intramolecular disulfide bond formed by Cys 242 and Cys 243 , were found to be common features in fungal GH30-7 xylanases used for phylogenetic analysis, whereas they were not conserved in GH30-8 enzymes, such as EcXynA from Gram-negative D. chrysanthemi (PDB code 2Y24) (9) and BsXynC from Gram-positive B. subtilis (PDB code 3KL0) (10) (Figs. 1 and 7 (B and C)). It should be noted that the ␤2-␣2 loop and the ␤-sheet structure

Characterization and crystal structure of Xyn30B
contribute to the formation of subsites Ϫ2a and Ϫ2b, respectively, which are involved in the substrate recognition (Fig. 8, E  and G). Alteration of Ϫ2a and Ϫ2b in GH30-7 provides a clue to understanding the catalytic diversity of Xyn30B.
Xyn30B sufficiently acts on linear XOS, whereas the specific activity of EcXynA for linear XOS is 3 orders of magnitude lower than that for aldouronic acid (9). The dual activity observed in Xyn30B is perhaps due to the contribution of the ␤2-␣2 loop, which recognizes the nonreducing end of XOS, especially the Asn 93 residue (Fig. 8, G and H). The subsite Ϫ3 of Xyn30B seems to be located at a different position from that of EcXynA with the presence of the loop; a candidate cleft has limited space, as shown by the dashed box in Fig. 8I. Our results indicate that the acidic XOS products and Xyl 4 are degraded in a xylobiohydrolase manner, despite them being of adequate lengths to bind to the Ϫ3 subsite (Fig. 6, A and B). These facts suggest that introduction of the xylose residue into the Ϫ3 subsite is likely interrupted by the limited space of subsite Ϫ3.  Fig. S6. The optimal tree with the sum of branch length ϭ 8.49592807 is shown. The tree is drawn to scale, with branch lengths in the same units, indicative of the evolutionary distances. The evolutionary distances were computed using the Poisson correction method (58) and are expressed as the number of amino acid substitutions per site. All positions containing gaps and missing data were eliminated. There were a total of 300 positions in the final data set. Evolutionary analyses were conducted in MEGA7 (59).

Characterization and crystal structure of Xyn30B
In contrast, Xyn30B shows obvious endo-glucuronoxylanase activity for substrates with the MeGlcA side chain recognized at the Ϫ2b subsite (Fig. 8E). Such substrates must bind to the Ϫ3 subsite and further downstream. R46A was found to prefer the xylobiohydrolase activity rather than the glucuronoxylanase activity (Fig. 6C), indicating that disturbance of the interaction between the MeGlcA side chain and the Ϫ2b subsite decreased endo-glucuronoxylanase activity without influencing the xylobiohydrolase activity. Based on these considerations, we speculate that the interaction between the MeGlcA side chain and the Ϫ2b subsite plays an important role in orientating the xylan main chain to introduce the xylose residue into the Ϫ3 subsite. Further structural studies using enzyme-substrate complexes are necessary to understand the bifunctionality of Xyn30B.
Some studies have shown that the loop-like roof structure is important factor for determining whether exo-or endo-hydrolysis activity occurs. Proctor et al. (31) have converted the enzyme specificity of Cellvibrio japonicus GH43 exo-arabinanase (CjArb43A) to an endo-type enzyme by removing a steric interaction at the nonreducing end of substrates. Santos et al. (32) have clarified that the steric interaction between the long loop and an arabinose residue at the reducing end of arabinan is important for the exo-action of a GH43 arabinanase isolated from rumen metagenome. The roof structures and its interaction with substrate at reducing or nonreducing ends of substrates have been reported to be important for other GH enzymes, such as GH8, -26, -46, and -74 (33)(34)(35)(36). These reports support our suggestion that the ␤2-␣2 loop of Xyn30B is critical for its exo-xylobiohydrolase activity, because an additional region, including Asn 93 (Ser 90 -Leu 94 ) in the ␤2-␣2 loop, appears to form a partial roof structure (Figs. 1 and 8I and Fig.  S5). This region is specific in Xyn30B and is not seen in XYN VI. These differences could explain why significant exo-xylanase activity is detected in Xyn30B.

Plasmid construction and fungal transformation
The plasmid pANC202 (24), which contains the pyrF gene and the glucoamylase (glaA) promoter and terminator regions, was used to construct the plasmids pANC215 and pANC281, which were used to express recombinant Xyn30B and Xyn30B R46A, respectively. Escherichia coli DH5␣ (Takara Bio, Kyoto, Japan) were used for the DNA procedures. The primers for the genomic region encoding xyn30B were designed based on the genome sequence of T. cellulolyticus registered in DDBJ/ EMBL/GenBank TM (DF933814.1) (39). The xyn30B gene was amplified using the forward primer 5Ј-ATTGTTAACA-GAATGGTGTTCAGCAAAGTCGCCG (with the HpaI site underlined), and the reverse primer 5Ј-AATCCTGCAGGT-CACTCGCACTCTGTAACAAAGCTTG (with the SbfI site underlined). The expression plasmid, pANC215, was constructed by ligating the xyn30B fragment that had been digested with HpaI/SbfI into the EcoRV/SbfI site of pANC202. The expression plasmid, pANC281, was constructed by site-directed mutagenesis of pANC215 using the KOD-plus-Mutagenesis kit (Toyobo, Osaka, Japan). The forward primer 5Ј-GCAGCGGAGGATATCTTCGGCAAGTACGGC (mutation site underlined) and the reverse primer 5Ј-TTGGAATGC-CTGTGAGCAGCCAAAG were used for PCR. The presence of all ligated gene fragments and locations was verified by DNA sequencing.
The plasmids pANC215 and pANC281 were transformed into protoplasts of T. cellulolyticus YP-4 by nonhomologous integration into the host chromosomal DNA (38). The strains producing recombinant Xyn30B and Xyn30B R46A were selected based on the amount of recombinant protein in culture supernatant as visualized by SDS-PAGE using NuPage 4 -12% Bis-Tris gels (Invitrogen) and were designated as Y215 and Y281, respectively.

Purification of Xyn30B and Xyn30B R46A
Purification of Xyn30B and Xyn30B R46A from culture supernatants of Y215 and Y281 strains, respectively, was performed using an ÄKTA purifier chromatography system (GE Healthcare, Buckinghamshire, UK) at room temperature. Culture supernatants were filtered through a 0.22-m polyethersulfone membrane and desalted using a HiPrep 26/10 desalting column (GE Healthcare) that had been equilibrated with 20 mM MES (pH 6.5). The desalted sample was applied to a Resource Q anion-exchange column (6 ml; GE Healthcare) that had been equilibrated with the same buffer, and protein peaks were eluted with a linear gradient of 0 -0.5 M NaCl (20 column volumes) at a flow rate of 4 ml min Ϫ1 . Fractions containing the target proteins were confirmed by SDS-PAGE and pooled. (NH 4 ) 2 SO 4 was added to a final concentration of 1.3 M, and then the samples were subjected to Source 15ISO (10 ml; GE Healthcare) hydrophobic interaction chromatography using a 1.3-0.7 M (NH 4 ) 2 SO 4 gradient (30 column volumes) in 20 mM sodium acetate buffer (pH 5.5) at a flow rate of 2.5 ml min Ϫ1 . The fractions containing target protein were pooled and were desalted and concentrated by ultrafiltration using Vivaspin 20-5K (Sartorius, Göttingen, Germany). The purified enzymes were preserved in a 20 mM sodium acetate buffer (pH 4.5) containing 0.01% NaN 3 at 4°C. Protein concentration was determined with a bicinchoninic acid protein assay kit (Thermo Scientific, Rockford, IL) using BSA (Thermo Scientific) as the protein standard.

Enzyme characterization
Xylanase activity was measured in a reaction mixture containing purified Xyn30B and 10 mg ml Ϫ1 beechwood xylan Characterization and crystal structure of Xyn30B (Megazyme, Wicklow, Ireland) in 50 mM sodium acetate (pH 4.0) at 40°C for 15 min. The reducing sugars from depolymerization of the substrate were measured using the DNS method (40). One unit of enzyme activity was defined as the amount of enzyme that catalyzed the formation of 1 mol of reducing sugar/min.
The optimal pH values and pH stabilities were examined using McIlvaine buffer for pH adjustment (41). To determine the optimal pH values, the reaction mixtures from pH 2.0 to 6.5 were incubated at 40°C for 15 min. To examine the pH stabilities, the enzymes were preincubated in buffer at pH values ranging from 2.0 to 7.0 for 30 min at 40°C, and the residual activity was subsequently measured under standard assay conditions using 10 mg ml Ϫ1 beechwood xylan. The optimal reaction temperature was examined at 35-60°C for 15 min in 50 mM sodium acetate (pH 4.0). To evaluate thermal stability, enzyme was preincubated in 50 mM sodium acetate (pH 4.0) at 4 -60°C for 30 min or 24 h, and then the residual activity was measured under standard assay conditions. To investigate the substrate specificity of Xyn30B, reaction mixture containing 10 mg ml Ϫ1 substrate was incubated at 40°C in 50 mM sodium acetate (pH 4.0). The following substrates were used: birchwood xylan, beechwood xylan, wheat arabinoxylan (Megazyme), carboxymethyl cellulose (Sigma-Aldrich), konjac glucomannan (Megazyme), and xyloglucan (Megazyme).
Xylobiohydrolase and transglycosylation activities were measured in a reaction mixture containing purified Xyn30B and 4 mM xylotriose (Xyl 3 , Megazyme) in 50 mM sodium acetate (pH 4.0) at 40°C for 15 min. The released products were determined by HPAEC-PAD analysis. One unit of xylobiohydrolase and transglycosylation activities was defined as the amount of enzyme that catalyzed the release of 1 mol of xylose and Xyl 5 per minute, respectively.

HPAEC-PAD analysis of hydrolysis reaction mixtures
HPAEC-PAD analysis of linear and acidic XOS hydrolysate was performed using a Dionex ICS-3000 ion chromatography system equipped with a CarboPac PA1 (Dionex, Sunnyvale, CA).
Analysis of acidic XOS was conducted at a flow rate of 1 ml min Ϫ1 as follows: (i) the system was equilibrated with 10 mM sodium hydroxide; (ii) after sample injection, 10 mM sodium hydroxide was run through the column for 3 min; (iii) a linear gradient of sodium hydroxide (10 -100 mM) was run for 7 min; (iv) a linear gradient of sodium acetate (0 -200 mM) in 100 mM sodium hydroxide was run for 40 min. The column was washed with 100 mM sodium hydroxide for 10 min after each sample analysis.

Mass spectrometry
The molecular mass of the purified Xyn30B was evaluated by MALDI-TOF MS with a Spiral TOF JMS-S3000 (JEOL, Tokyo, Japan). The purified sample was applied to the MALDI target plate after dilution into a mixture containing 0.5% (w/v) sinapinic acid, 0.1% TFA, and 25% acetonitrile. Monovalent and bivalent ions from conalbumin (75 kDa) included in the Gel Filtration Calibration Kit HMW (GE Healthcare) were used for external mass calibration. Instrument control, data acquisition, and data processing of all experiments were achieved using MSTornado (JEOL).
Electrospray ionization single-stage and multistage MS in the negative ion mode (ESI(Ϫ)-MS n ; n ϭ 1-3) were used for the molecular and structural analyses of acidic XOS using an ama-Zon SL-STT2 ion trap (Bruker, Bremen, Germany). Samples were diluted in methanol (MeOH) and introduced into the ionization source in infusion mode using a syringe pump at a flow rate of 10 l min Ϫ1 . The apparatus was operated in enhanced resolution mode (mass range: 50 -2200 m/z, scanning rate: 8,100 m/z per second). In MS n experiments (n Ͼ 1), the width of the selection window was set at 1 Da to obtain clean isotopic selection. The amplification of the excitation was set according to the experiment to reach a survival yield (abundance of the precursor ion divided by the sum of the product and precursor ion abundances) at ϳ20%. Instrument control, data acquisition, and data processing of all experiments were achieved using Compass 1.3 SR2 (Bruker), whereas mMass 5.5.0.0 (44) was used for data treatment and artworks.

X-ray crystallography
Purified Xyn30B was concentrated to 30 mg ml Ϫ1 for crystallization. Crystals were obtained with the hanging-drop vaper diffusion method at 20°C for a week. The drop was composed of 1.5 l of protein solution mixed with 1.5 l of reservoir solution containing 30% PEG 4000, 0.1 M Tris-HCl (pH 8.1), 200 mM sodium acetate and equilibrated against 500 l of reservoir solution.
The Xyn30B crystal was soaked with the reservoir solution supplemented with 30% glycerol as a cryoprotectant and then flash-cooled in liquid nitrogen. X-ray diffraction data of the Characterization and crystal structure of Xyn30B crystal were collected to 2.25 Å resolution at 100 K at the SPring-8 beamline BL44XU. Diffraction images were checked with adxv (http://www.scripps.edu/tainer/arvai/adxv.html) 3 and integrated and scaled with XDS (version: January 26, 2018) (45). Phasing was performed using Molrep 11.6 in CCP4 7.0 (46,47) with CaXyn30A (PDB code 5CXP) as the model, which had been processed using Sculptor in Phenix 1.12 (48,49). The first model was refined using AutoBuild (50). The model was manually completed using Coot 0.8.9 and refined using Refmac 5.8 (51,52). Model quality was verified using MolProbity 4.4 (53). Molecular figures were generated with Open-source PyMol 1.8 (54). Secondary structures of Xyn30B, EcXynA, and BsXynC were assigned using STRIDE (55).
Author contributions-H. I. designed the study. H. I. and Y. N. prepared and characterized the enzymes. T. F. and S. I. collected the mass spectrometry data. Y. N. and M. W. conducted the X-ray crystallography experiments. A. M. and H. I. coordinated the study. All authors contributed to the writing of this manuscript and approved the final version.