Identification, characterization, and structural analyses of a fungal endo-β-1,2-glucanase reveal a new glycoside hydrolase family

endo-β-1,2-Glucanase (SGL) is an enzyme that hydrolyzes β-1,2-glucans, which play important physiological roles in some bacteria as a cyclic form. To date, no eukaryotic SGL has been identified. We purified an SGL from Talaromyces funiculosus (TfSGL), a soil fungus, to homogeneity and then cloned the complementary DNA encoding the enzyme. TfSGL shows no significant sequence similarity to any known glycoside hydrolase (GH) families, but shows significant similarity to certain eukaryotic proteins with unknown functions. The recombinant TfSGL (TfSGLr) specifically hydrolyzed linear and cyclic β-1,2-glucans to sophorose (Glc-β–1,2-Glc) as a main product. TfSGLr hydrolyzed reducing-end–modified β-1,2-gluco-oligosaccharides to release a sophoroside with the modified moiety. These results indicate that TfSGL is an endo-type enzyme that preferably releases sophorose from the reducing end of substrates. Stereochemical analysis demonstrated that TfSGL is an inverting enzyme. The overall structure of TfSGLr includes an (α/α)6 toroid fold. The substrate-binding mode was revealed by the structure of a Michaelis complex of an inactive TfSGLr mutant with a β-1,2-glucoheptasaccharide. Mutational analysis and action pattern analysis of β-1,2-gluco-oligosaccharide derivatives revealed an unprecedented catalytic mechanism for substrate hydrolysis. Glu-262 (general acid) indirectly protonates the anomeric oxygen at subsite −1 via the 3-hydroxy group of the Glc moiety at subsite +2, and Asp-446 (general base) activates the nucleophilic water via another water. TfSGLr is apparently different from a GH144 SGL in the reaction and substrate recognition mechanism based on structural comparison. Overall, we propose that TfSGL and closely-related enzymes can be classified into a new family, GH162.

Despite the important physiological functions of ␤-1,2-glucans in nature, ␤-1,2-glucan-associated enzymes and proteins have been explored much less than other GH enzymes that degrade polysaccharides such as ␤-1,3-glucan and cellulose (␤-1,4-glucan), due to the difficulty in practical preparation of cyclic ␤-1,2-glucans. This may be attributed to the fact that cyclic ␤-1,2-glucan synthase is a transmembrane enzyme and uses UDP-glucose, which is a more expensive starting material than glucosides such as sucrose and starch, as a substrate (17).

Purification of an SGL from fungi
Among fungi known to inducibly secrete SGLs by cyclic ␤-1,2-glucans, we examined the SGL activities in culture filtrates of T. funiculosus grown in the presence of linear ␤-1,2glucan as a sole carbon source. Because the culture filtrate of T. funiculosus degraded the linear ␤-1,2-glucan, we purified the secreted TfSGL by three chromatographies. Although the SGL fractions contained BGL activity during the purification steps, SGL activity was completely separated from BGL activity by size-exclusion chromatography (SEC), the final purification step ( Fig. 1A and Table 1), suggesting that the purified enzyme is an actual SGL. In Fig. 1A, DPs of the reaction products were somewhat obscure, which might be attributed to a slight smiling effect by NaCl in the enzyme solutions. The fractions containing the target enzyme gave a single band that migrated at ϳ60 kDa on SDS-PAGE, indicating that TfSGL had been successfully purified to homogeneity. On SEC analysis, TfSGL eluted at 59 kDa (Fig. S1), suggesting that TfSGL is a monomeric enzyme. The purified TfSGL is highly specific for ␤-1,2glucan among the polysaccharides examined (data not shown).

Glycosylated protein analysis of the purified enzyme
To determine whether TfSGL is glycosylated, periodic acid-Schiff stain analysis of the purified enzyme was performed. The enzyme was detected on both protein and sugar-chain staining, indicating that TfSGL is a glycoprotein (Fig. 1B). After the purified enzyme had been incubated with glycopeptidase F, which specifically cleaves N-glycans in glycoproteins, TfSGL was detected only on protein staining at ϳ50 kDa. Therefore, TfSGL is a glycoprotein possessing only N-glycan of ϳ10 kDa.

Analysis of action patterns on ␤-1,2-glucans and Sop n s
The action pattern of TfSGL on linear ␤-1,2-glucan was determined by TLC analysis. TfSGL hydrolyzed ␤-1,2-glucan to Sop 2 as a main product, implying that TfSGL preferably releases Sop 2 from either end of ␤-1,2-glucan ( Fig. 2A). In addition, Sop 4 was accumulated as a minor product. Sop n s with a DP of 5 or more were not visible in the TLC plate at a late stage of the reaction, although Sop 5 was temporarily accumulated at the middle stage of the reaction probably due to slow hydrolysis of Sop 5 by TfSGL ( Fig. 2A). TfSGL hydrolyzed Sop 5 to Sop 2 and Sop 3 but did not degrade Sop 3 or Sop 4 (Fig. 2B). These results indicate that TfSGL hydrolyzes Sop n s with DP of 5 or more to Sop 2 as a main product.

Sequence analysis
Because identification of the internal amino acid sequence of the purified enzyme was unsuccessful by LC-tandem MS analysis, N-terminal and internal peptide sequencing was per-A novel ␤-1,2-glucanase from fungi formed. The obtained N-terminal and two internal peptide sequences are shown in Fig. 3. Then, we cloned a region containing the whole TfSGL gene of genomic DNA (gDNA) and complementary DNA (cDNA) (assigned accession number by DNA Data Bank of Japan, LC430902). The predicted transcription start site and the predicted polyadenylation signal sequence were found ϳ100 bp upstream of the initiation codon and ϳ300 bp downstream of the termination codon, respectively. The N-terminal signal peptide in TfSGL was predicted to have 18 residues, whereas the actual N terminus of TfSGL is the 22nd residue. Several plausible N-glycosylation sites predicted in TfSGL are asparagine residues (residues 41, 60, 120, 143, 194, 197, 237, 361, and 384), with this being consistent with the results of glycosylated protein analysis of TfSGL. Prediction of GPI-modification sites in TfSGL was performed because some TfSGL homologs are annotated as GPI-anchored proteins. However, no predicted GPI-modification site was found in TfSGL or these homologs. This is consistent with the fact that internal peptide 2 was quite near the C terminus.   [2][3][4][5] . The numbers beside the TLC plates represent DP of Sop n s. The purified TfSGL (1.9 and 3.8 g/ml for hydrolytic reactions of ␤-1,2-glucan and Sop n s, respectively) was incubated in 100 mM acetate-Na buffer (pH 5.0) containing 0.2% ␤-1,2-glucan or 5 mM Sop n s with DP of 3-5 at 20°C. Arrows represent ␤-1,2-glucan used for reactions. The origins of the TLC plates are shown as horizontal lines denoted by asterisks. A novel ␤-1,2-glucanase from fungi

Phylogenetic analysis
We constructed a phylogenetic tree, as shown in Fig. 4. The phylogenetic tree comprised only proteins annotated as hypothetical, unnamed, or GPI-anchored proteins. Although several homologs are annotated as glycosyl hydrolase family proteins, the annotations depend on the GH1 domain fused with regions homologous to TfSGL, and no annotation is given to the homologous regions. The biochemical functions of these TfSGL homologs are unknown, and they exhibit no sequence similarity to any known GH families. In addition, even structural prediction with InterPro (https://www.ebi.ac.uk/interpro/) 3 did not allow classification of TfSGL into any of the families. Almost all of the homologs are from Eukaryotes with some exceptions, such as a homolog from Elusimicrobia bacterium, which was found from low O 2 conditions as an environmental uncultured species (Fig.  4). No TfSGL homolog was found in Talaromyces marneffei or Talaromyces stipitatus, strains used for a Mascot search in the LC-tandem MS analysis.

General properties of TfSGLr
TfSGLr was successfully produced by Pichia pastoris. The purified enzyme migrated as a single band with a similar molecular mass to that of the native TfSGL on SDS-PAGE (Fig. S2A). Because TfSGLr showed hydrolytic activity toward ␤-1,2-glucan like the native TfSGL, ␤-1,2-glucan was used as a substrate to investigate the pH and temperature profiles. TfSGLr was stable at pH 4.0 -7.0 and up to 30°C. TfSGLr exhibits the highest activity at 60°C and at pH 4.0 -4.5 (Fig. S2, B and C).

Substrate specificity of TfSGLr
To determine the substrate specificity of the TfSGLr, hydrolytic activity toward various polysaccharides was examined. The enzyme acted on linear ␤-1,2-glucan specifically (specific activity was 17 units/mg as Sop 2 -releasing activity). Then, the chain-length specificity of the enzyme for Sop n s was examined. The enzyme hydrolyzed Sop 5 to Sop 2 and Sop 3 , and Sop 6 to Sop 2 and Sop 4 as main products, with Sop 3 as a minor product, and Sop 7 to equal amounts of Sop 2-5 . The enzyme did not hydrolyze Sop 3 or Sop 4 (Fig. 5A). These catalytic properties are essentially the same as those of the native TfSGL. The hydrolytic velocity toward Sop 5 was ϳ5 times lower than that toward Sop 6 . These results suggest that TfSGLr hydrolyzes ␤-1,2-glucan with a DP of 5 or more but requires DP of at least 6 for efficient hydrolysis.

Action patterns of TfSGLr
TfSGLr hydrolyzed both cyclic and linear ␤-1,2-glucans at similar velocity, indicating that TfSGLr is an endo-type enzyme The first (G) and 7th (X) residue in the internal peptide 2 is a presumed and an undetermined residue, respectively. Only the 6th residue in the N-terminal peptide and the 7th residue in the internal peptide 2 were replaced with cysteine residues in the deduced amino acid sequence of TfSGL. The differences in the sequences are attributed to the fact that cysteine cannot generally be detected unless it is pyridylethylated. The degenerate and specific PCR primers used for cloning are represented with thin and bold arrows, respectively. Primer pairs used for degenerate PCR and specific PCR are boxed with dotted and solid lines, respectively. The numbers above the coding DNA sequence boxes and beside the primers represent the nucleotide numbers from the start codon on gDNA.
A novel ␤-1,2-glucanase from fungi (Fig. 5B). A main product released from linear ␤-1,2-glucan was Sop 2 . Sop 3-5 were also produced as minor products after a 1-h reaction. To determine which ends of the linear ␤-1,2-glucan are hydrolyzed by TfSGLr, the action patterns on Sop n s modified at the reducing-end by NaBH 4 (rSop n s) were investigated. Although Sop 2 was not observed at the early stage of the reaction, rSop 3 was found to be released from rSop 6 -8 despite the modification of the reducing end (Fig. 5C). Considering that rSop 2 was not produced, rSop 3 is likely released as a disaccharide by TfSGLr, probably because the modified moiety adopting an open chain form is not recognized by TfSGLr. In addition, there was no difference in hydrolytic velocity for the substrates regardless of the modification at the reducing end of the substrates (Fig. 5, A and C), suggesting that modification of the substrates does not affect reaction velocity. These results suggest that TfSGL is an endo-type enzyme but quite preferably releases Sop 2 from the reducing end of ␤-1,2-glucan. This is a different property from that of PdSGL, which releases Sop 2 from the nonreducing end of linear ␤-1,2-glucan (26).

Stereochemistry of hydrolysis catalyzed by TfSGLr
To determine the reaction mechanism of TfSGLr, the anomeric configurations of hydrolysates released from ␤-1,2-glucan were examined by 1 H NMR. The ␤-anomer signal derived from the anomeric axial proton at the reducing end in liberated Sop n s (H1␤ R , around ␦ 4.7) overlapped with chemical shifts derived from water and anomeric axial protons in internal glucose (Glc) moieties. Thus, the signals derived from the anomeric proton of the ␣-anomer of the released Sop n s (H1␣ R , around ␦ 5.4) and the C2 proton at the nonreducing end of all the reaction products (H2 NR , around ␦ 3.3) were used for stereochemical analysis as described by Abe et al. (25). Both peak areas increased immediately after addition of the enzyme (Fig.  7, A and B). Then, the former peak area gradually decreased during the reaction due to the nonenzymatic mutarotation of the products, whereas the latter gradually increased probably due to slow hydrolysis of Sop 5 , which was accumulated as shown in Fig. 5B. This result suggests that TfSGL is an inverting enzyme. We also examined the change of the degree of optical rotation during the hydrolysis of ␤-1,2-glucan by TfSGLr. An increase in the degree of optical rotation at the early stage of the reaction and a dramatic decrease in it in addition to aqueous ammonia was observed (Fig. 7C), with this being the same pattern as that of CpSGL, an inverting enzyme (25). This result also supports an inverting mechanism of TfSGL.

Overall structure of TfSGLr
The apo structure of TfSGLr was determined at 2.0 Å resolution using the iodide single-wavelength anomalous diffraction-phasing method (Table S1 and Fig. 8). There are two molecules in an asymmetric unit, and their root mean square deviation (RMSD) value was determined to be 0.1 Å. The struc- Asterisks represent homologs whose sequences were used for sequence alignment with TfSGL in Fig. 10.
A novel ␤-1,2-glucanase from fungi ture consists of mainly antiparallel ␣-helices forming inner and outer rings, indicating that TfSGLr folds into a single (␣/␣) 6barrel domain (Fig. 8). A structural homology search using the Dali server (31) showed that the overall structure of TfSGLr is similar to those of GH144 family members, including CpSGL (Protein Data Bank (PDB) code 5GZH) (Table S2). However, TfSGLr shows remarkably low amino acid sequence identities with GH144 proteins according to structure-based alignment  3,5,7,9,11 . The substrates used for the reactions are shown above the TLC plates. Each substrate (0.2%) was hydrolyzed with TfSGLr (19.3 g/ml for hydrolysis of Sop 5 and rSop 6 and 3.2 g/ml for the other substrates). The origins of the TLC plates are shown as horizontal lines denoted by asterisks. The numbers beside the TLC plates represent DP of Sop n s. B, arrows indicate ␤-1,2-glucans used for reactions. C, numbers with the letter r beside the TLC plates represent rSop n s. The pattern diagrams of rSop n s are shown below the TLC plates. The open and closed circles in the pattern diagrams represent Glc and the reduced Glc moieties, respectively. Cleavage sites are represented as arrowheads. A novel ␤-1,2-glucanase from fungi with the PDBeFold server (less than 11%) (32,33). TfSGLr also exhibits structural similarities to other GH family enzymes, a GH126 ␣-amylase from Clostridium perfringens (PDB code 3REN) (34), and a GH15 glucoamylase from Saccharomycopsis fibuligera (PDB code 1AYX). The similarities are lower than those of the GH144 enzymes according to Z-scores (RMSD values) (Table S2). TfSGLr has a cleft crossing the surface of the structure, and there is a large pocket at the center of the cleft (Fig. 8C). The pocket is located at the center of the (␣/␣) 6 -barrel as in most (␣/␣) 6 -barrel GH enzymes.

Complex structure with Sop 2 and Glc
In the obtained complex structure with Sop 2 and Glc by soaking a TfSGLr crystal in Sop 2 , the electron density of a Sop 2 molecule was clearly observed at the center of the cleft (Fig. 9A). Considering that TfSGLr produces mainly Sop 2 from the reducing end of linear ␤-1,2-glucan, the Sop 2 molecule seems to be located at subsites ϩ1 to ϩ2. This is supported by the TfS-GLr structure in complex with Sop 7 , as described below.
The Glc moiety at the nonreducing end of the Sop 2 molecule forms hydrogen bonds with Trp-155 and Asp-177 (Fig. 9A).
The hydrophobic side chain of Leu-176 interacts with the pyranose ring of the Glc moiety. In addition, the aromatic ring of Trp-169 undergoes a hydrophobic interaction with the C6 atom of the Glc moiety. The Glc moiety at the reducing end of the Sop 2 molecule forms six hydrogen bonds with Asp-259, Glu-262, and His-316. The side chain of His-316 adopts alternative conformations in the apo TfSGLr, unlike in the case of the Sop 2 complex (Fig. 9A). Two aromatic residues (Tyr-311 and Trp-312) undergo hydrophobic stacking interactions with C6 and the pyranose ring in the Glc moiety, respectively.
A relatively weak electron density that can be fitted with a Glc molecule was also observed in the pocket. The Glc molecule forms hydrogen bonds with Asp-72, Lys-94, Asn-360, and Asp-446 (Fig. 9A). The aromatic rings of Trp-155 and Tyr-373 undergo hydrophobic stacking interactions with the pyranose ring of the Glc molecule.

Determination of subsite positions in the complex structure with Sop 7
To elucidate a more detailed substrate recognition mechanism by obtaining the structure of the Michaelis complex of A novel ␤-1,2-glucanase from fungi TfSGLr, we searched for an inactive TfSGLr mutant. Because mutation of Glu-262 in the substrate-binding pocket to glutamine abolished the activity of TfSGL toward ␤-1,2-glucan, we soaked a crystal of the E262Q mutant in ␤-1,2-glucan. In each catalytic pocket of two molecules (chain A or B) in the asymmetric unit of the complex, the electron density of a Sop 5 or Sop 7 molecule was observed, respectively. Therefore, chain B was used for description of the complex. Among the Glc moieties of the Sop 7 molecule, the electron density of a Sop 6 moiety was clearly observed (Fig. 9B), whereas that of the remaining Glc moiety was ambiguous.
It should be noted that the fourth Glc moiety from the nonreducing end of the Sop 7 molecule forms a skew-boat conformation ( 1 S 3 ) (Fig. 9C) according to Cremer-Pople parameters ( (°), (°), and Q (Å) of the Glc moiety are 204.341, 81.202, and 0.713, respectively) (35). In contrast, the other Glc moieties form a 4 C 1 conformation. In GH5 retaining endo-␤-1,4-glucanase from Bacillus agaradhaerens and GH63 inverting ␣-glycosidase from Escherichia coli, the Glc moieties of their substrates at subsite Ϫ1 adopt a 1 S 3 conformation (36,37). This twisted conformation enables a nucleophile to be located where nucleophilic attack to an anomeric carbon is possible. In TfSGLr, a water molecule is located (3.1 Å) near the anomeric carbon of the Glc moiety with a 1 S 3 conformation. The angle formed by this water, the anomeric carbon, and an oxygen atom of the glycosidic bond between subsites Ϫ1 and ϩ1 is 161.1°, which is suitable for nucleophilic attack (in-line-attack) on the anomeric carbon. These observations strongly suggest that the position of the fourth Glc moiety is subsite Ϫ1 and that TfSGL accommodates the Sop 7 molecule at subsites Ϫ4 to ϩ3, a Michaelis complex being formed (Fig. 9B).

Substrate-binding mode in the Sop 7 complex
The Glc moieties of Sop 7 at subsites ϩ1 and ϩ2 well overlapped those of the Sop 2 molecule in the TfSGLr-Sop 2 complex, and this is consistent with the presumed subsite positions of the Sop 2 molecule in the TfSGLr-Sop 2 complex described above (Fig. 9D). The Glc molecule in the TfSGLr-Sop 2 complex is also well-superimposed with the Glc moiety at subsite Ϫ3 in the TfSGLr-Sop 7 complex. Thus, the Glc moieties of Sop 7 at subsites Ϫ3, ϩ1, and ϩ2 are firmly recognized in almost the same way as in the Sop 2 complex (Fig. 9, A and B). The Glc moieties at subsites Ϫ3 to ϩ2 are apparently accommodated in the large substrate pocket, which looks like a "gravy boat" (Fig. 9D).
Unlike subsites Ϫ3, ϩ1, and ϩ2, subsites Ϫ2 and Ϫ1 seem to be unfavorable factors for substrate binding. The Glc moiety at subsite Ϫ2 forms only one hydrogen bond with Ser-375. The twisted glycosidic bond between subsites Ϫ1 and ϩ1 (see below), and the twisted conformation of the Glc moiety at subsite Ϫ1 also make the binding unstable. These observations imply the importance of subsite Ϫ3 for substrate binding on the minus subsite side. Taken together with the requirement of subsite ϩ2 for catalysis, as described later, structural observations suggest that at least a DP of 5 is needed for Sop n to act as a substrate. This is consistent with the finding that Sop 5 is a minimum substrate of TfSGLr (Fig. 5A).
The 2-and 1-hydroxy groups of the Glc moieties at subsites Ϫ4 and ϩ3, respectively, face the solvent (Fig. 9D), indicating that TfSGL can accommodate a substrate extending beyond both subsites Ϫ4 and ϩ3. This observation is consistent with the endolytic property of TfSGL.
At subsite Ϫ4, only a hydrogen bond with Trp-169 and a hydrophobic stacking interaction with the aromatic ring of Trp-155 were found, whereas no interaction was observed at subsite ϩ3. This difference is consistent with the observed action patterns that Sop 6 was preferentially hydrolyzed to Sop 4 and Sop 2 . Judging from these observations, the binding mode of Sop 7 is consistent with the characteristics of TfSGL, suggesting that binding of Sop 7 is productive.
Most of the substrate recognition residues are conserved among TfSGL homologs except that two residues in TfSGLr (Tyr-311 and Ser-375) exhibit a little variety (Fig. 10). This implies that the function and structure of TfSGL are conserved among TfSGL homologs.

Comparison of conformations of ligands between ␤-1,2-glucan-associated proteins
The conformations of Sop [3][4][5] have been reported as ligands in the complex structures of LiSO-BP (22). In the complexes, the Sop 3-5 molecules adopt stable conformations, and the Glc moieties in the ligands line up in a zigzag manner. Compared with these ligands, a feature of the Sop 7 molecule in TfSGLr is a twisted ( 1 S 3 ) conformation at subsite Ϫ1, as described above. It A, front view of the overall apo TfSGLr structure (left) and the structure rotated by 90°around the y axis (right). The ␣-helices, 3 10 -helices, ␤-strands, and loop are shown in green, cyan, yellow, and blue, respectively. Four asparagine residues (Asn-143, Asn-237, Asn-361, and Asn-384) bonded with GlcNAc, and the GlcNAc moieties and residues constituting three disulfide bonds (Cys-27-Cys-516, Cys-230 -Cys-251, and Cys-345-Cys-499) are represented by magenta, orange, and white sticks, respectively. B, order of helices constituting an (␣/␣) 6 -barrel in TfSGLr. The barrel is represented by a rainbow cartoon. The helices in TfSGLr are numbered in order from the N terminus. C, surface of the overall structure of apo TfSGLr.
A novel ␤-1,2-glucanase from fungi is observed that the anomeric hydroxy group in the Glc moiety is located at a lifted position when pair-fitting of the pyranose rings of substrates at subsites Ϫ3 to Ϫ1 in TfSGLr and at units A to C (the Sop 3 moiety from the nonreducing end of Sop 5 ) in LiSO-BP is performed (Fig. S3). The Glc moiety of subsite ϩ1 in the TfSGLr-Sop 7 complex is rotated by ϳ100°against that in the LiBGL-Sop 2 complex, suggesting that the relative positions of subsite ϩ1 against subsite Ϫ1 are quite different between the two enzymes ( Fig. S3). The difference in the position of subsite ϩ1 makes subsites Ϫ2 to ϩ1 crowded in the TfSGLr-Sop 7 complex (Fig. 9B). A "gravy boat"-like large pocket might be required to accommodate substrates with the twisted conformation (Fig. 9D). Unlike TfSGLr, Sop n s are sandwiched by the closure motion of two domains in LiSO-BP, and LiBGL has a narrow "coin slot"-like structure at subsite ϩ1 (a pyranose ring is sandwiched by two aromatic residues) (21,22). In the case of SOGP from Lachnoclostridium phytofermentans, a Glc moiety at subsite ϩ1 is flipped at the position corresponding to subsite ϩ1 in LiBGL (21,24). The structure of TfSGLr exhibits a novel substrate-binding mode among ␤-1,2glucan-associated enzymes and proteins.

Mutational analysis of candidates for a general base
In general, an inverting GH enzyme hydrolyzes a glycosidic bond in its substrate through a single-displacement mechanism using two acidic residues (Fig. 11A). A general base activates a water for nucleophilic attack on the anomeric carbon, whereas a general acid protonates the glycosidic bond oxygen directly (38). However, no acidic residue directly interacts with the nucleophilic water molecule or is located within the proton transfer distance range from the glycosidic bond oxygen atom at the cleavage site in the case of TfSGLr (Fig. 11B). Therefore, TfSGLr unlikely utilizes a canonical reaction mechanism.
As to the reaction pathway for a general base, only Tyr-373 interacts with the nucleophilic water among the amino acid residues (Fig. 11B). Furthermore, there is no acidic residue hydrogen-bonded with Tyr-373. Therefore, we added the following residues that interact indirectly with the nucleophilic water via multiple waters as candidate residues for catalysis: His-316, Ser-358, Tyr-396, Glu-430, Thr-444, and Asp-446. Glu-180 was also added as a candidate because it is an acidic residue located near subsite Ϫ1 (Fig. 11B). These candidate residues, except Thr-444, are highly conserved among TfSGL homologs (Fig. 10). Therefore, mutational analysis of all these residues was carried out for identification of the general base, although the T444A mutant was not expressed as a soluble protein. The D446N mutant showed no activity toward ␤-1,2glucan on colorimetric assay or TLC analysis ( Fig. 11C and Table 2). The relative activity of the E430A mutant was less than 1%, which could also be regarded as the activity level of a catalytic mutant. Even when the 3-methyl-2-benzothiazolinone hydrazone (MBTH) method (39), the most sensitive method for quantification of Sop n s, was used for assay, the relative activity of the E430A mutant did not show a change (Table 2). However, it is difficult to regard Glu-430 as a general base, because there needs to be three additional waters to reach the nucleophilic water, and one of the hydrogen bond lengths between these waters (3.6 Å) is longer than a general one. Although there is also a proton network from Glu-430 to the nucleophilic water A novel ␤-1,2-glucanase from fungi Figure 10. Multiple sequence alignment of TfSGL and its homologs. Multiple alignment using TfSGL homologs with at least 37% sequence identity was carried out. The homologs are represented as GenBank TM or NCBI reference sequence accession numbers. The symbols below the sequences are represented as follows: open circles, candidates for catalytic residues of TfSGLr; closed circles, residues involved in substrate recognition; closed triangles, N-glycosylated asparagine residues; and closed stars, the general acid (Glu-262) and the general base (Asp-446) of TfSGL. The same numbers below sequences represent disulfide bond pairs. Residue numbers above sequences are based on the amino acid sequence of the native TfSGL. The secondary structures of TfSGLr are shown above the sequence. The order of helices constituting (␣/␣) 6 barrels in TfSGLr is shown in parentheses.
A novel ␤-1,2-glucanase from fungi via Tyr-373, proton transfer unlikely takes place via this route. This is because the Y373F mutant retained too high hydrolytic activity toward ␤-1,2-glucan as a residue involved in the reaction pathway ( Table 2). In addition, the hydrolytic activity of the E430A mutant detected by TLC analysis was apparently higher than those of D177N and E262Q mutants (the candidates for a general acid described below) and the D446N mutant (Fig.  11C). The other candidates (His-316, Ser-358, and Tyr-396) are also unlikely general bases. This is because the mutants (H316Q, S358A, and Y396F) retained sufficient hydrolytic activity toward ␤-1,2-glucan, although the H316A mutant showed remarkably decreased activity. These results suggest that Asp-446 probably acts as a general base via two water molecules.
Because Tyr-373 interacting with the nucleophilic water is not located between Asp-446 and the nucleophilic water, Tyr-373 probably stabilizes the position of the nucleophilic water. The drastically decreased activities of the H316A and E430A mutants might imply the importance of the water network for catalytic efficiency.

Mutational analysis of candidates for a general acid
As to the reaction pathway for a general acid, only 3-hydroxy groups in the Glc moieties at subsites ϩ1 and ϩ2 are found within the proton transfer distance range from the oxygen atom in the cleavage site. The 3-hydroxy groups at subsites ϩ1 and ϩ2 form hydrogen bonds with Asp-177 and Glu-262, respectively. Although Asp-177 and Glu-262 are located too far from the glycosidic bond's oxygen atom for direct proton transfer (distances of 4.8 and 4.4 Å, respectively), these residues are highly conserved among TfSGL homologs (Fig. 10). Therefore, it is predicted that Asp-177 and/or Glu-262 act as a general acid via the hydroxy groups of the Glc moieties. TLC analysis  a Specific activity of TfSGLr was calculated by measuring the amount of Sop 2 produced from ␤-1,2-glucan using the GOPOD method. Specific activity and relative activity calculated by measuring the amount of produced Sop n s using the MBTH method are shown in parentheses. b Specific activities of wildtype TfSGLr calculated by the GOPOD and MBTH methods were defined as 100% relative activities. c ND represents the data of less than 0.02 units/mg (0.01% relative activity).
A novel ␤-1,2-glucanase from fungi showed that no hydrolytic products derived from linear ␤-1,2glucan by the E262Q mutant were detected, whereas Sop 2 was slowly produced by the D177N mutant (Fig. 11C). However, according to the colorimetric assay, the activities of both the D177N and E262Q mutants toward ␤-1,2-glucan were undetectable (Table 2). This finding suggests that both Asp-177 and Glu-262 still could be catalytic residues. Therefore, the action patterns of TfSGLr on Sop n s with the appropriate 3-hydroxy groups deoxygenated were analyzed as described below.

Action pattern analysis of 3-deoxy Sop n -derivatives
To determine the actual general acid of TfSGL, the action patterns for Sop 5-6 deoxygenated at their 3-hydroxy groups at the first or second Glc moiety from the reducing end (3dSop [5][6] and 3ЈdSop 5 , respectively) were investigated (Fig. 12). 3ЈdSop 2 was released from 3ЈdSop 5 by TfSGLr, whereas 3dSop 5 was hardly hydrolyzed even in the presence of an excessive amount of TfSGLr (Fig. 12, A and B). We also examined the action pattern for 3dSop 6 , a substrate with a preferable chain length. As a result, 3dSop 3 and Sop 3 were released from 3dSop 6 without production of 3dSop 2 and Sop 4 (Fig. 12C). This finding indicates that deoxygenation of the 3-hydroxy group of the Glc moiety at subsite ϩ2 completely inhibits the cleavage of the substrate despite that the glycosidic bond between the second and third Glc moieties from the reducing end is the preferable cleavage site in Sop 6 ( Figs. 5 and 12D). These results exclusively suggest that Glu-262 acts as a general acid via the 3-hydroxy group of the Glc moiety at subsite ϩ2. A novel ␤-1,2-glucanase from fungi

Molecular dynamics simulation of TfSGLr
To exclude the possibility that the crystal structures of TfSGL are affected by crystal packing, molecular dynamics (MD) simulations of the apo structures (see "Experimental procedures") and the Sop 7 complex of TfSGLr were performed. Neither the TfSGLr structure nor the Sop 7 -complex structure significantly changed during the simulations with the average RMSD values of 0.8 -1.3 Å (Fig. S4). The C␣ root mean square fluctuation (RMSF) values of residues around the catalytic pocket of TfS-GLr were lower than 1.0 Å (Table S3 and Fig. S4). Principal component analysis (PCA) and cluster analysis demonstrated that a representative MD structure of the complex and its crystal structure are almost the same with the C␣ RMSD of 0.6 Å and that a representative MD structure of the apo state showed a slight opening movement of the catalytic pocket relative to the apo crystal structure (Fig. S5). These results suggest that crystal packing does not affect the structure of TfSGL in complex with Sop 7 .

Noncanonical reaction mechanism of TfSGL
The proposed catalytic reaction of TfSGL proceeds as follows. Glu-262 (general acid) indirectly protonates the anomeric oxygen at subsite Ϫ1 via the 3-hydroxy group of the Glc moiety at subsite ϩ2. Asp-446 (general base) indirectly activates the water for nucleophilic attack via another water (Fig. 13). A proton relay mechanism in TfSGL is called a "Grotthuss"-style mechanism (40) and is a noncanonical one in GH enzymes.
The pK a prediction of Glu-262 and Asp-446 was performed by the PROPKA3.0 (41) server using the protein moiety in the Sop 2 -complex structure. The pK a values of Glu-262 (8.22 and 8.21 in chain A and B, respectively) were clearly higher than those of Asp-446 (3.37 and 3.35, respectively), which is consistent with the proposed TfSGL reaction mechanism. The difference in the pK a values may be mainly attributed to a decrease in pK a value of Asp-446 by a positive charge of Lys-94 close to Asp-446 and an increase in pK a value of Glu-262 by negative charges of Asp-259 and Glu-317 close to Glu-262.
In GH130 4-O-␤-D-mannosyl-D-glucose phosphorylase from Bacteroides fragilis, the 3-hydroxy group in the substrate at subsite Ϫ1 is found within the range of hydrogen bond interaction with the acidic residue and the glycosidic bond oxygen (42). Based on the structural features and mutational analysis of the GH130 enzyme, it is postulated that the 3-hydroxy group plays a role as a mediator of protonation. This mechanism is the same as that of TfSGL in that a proton is relayed via a substrate hydroxy group. However, the two enzymes use different hydroxy groups.
Several retaining and inverting enzymes follow a Grotthuss mechanism, in which a general base remotely activates the nucleophilic water via another water molecule (GH6, GH101, and GH136) (43)(44)(45). For example, cellobiohydrolase from Trichoderma reesei (Cel6A) is an inverting GH and is considered to follow a Grotthuss mechanism. TfSGLr is similar to Cel6A in that another water molecule mediates proton transfer. However, the positions of general bases in the two enzymes are quite different.

Comparison of the TfSGLr and CpSGL structures
Because TfSGLr and CpSGL exhibit the same substrate specificity and show structural similarity despite their exclusively different amino acid sequences, we compared the overall structures of the two enzymes. The positions of helices constituting (␣/␣) 6 barrels are similar between them, suggesting a relationship in structural evolution (Fig. 14A) (15). Although the ligands in both enzymes deviated in the superimposition of the overall structures, these ligands are well-superimposed by fitting the corresponding Glc moieties between the two enzymes. The substrate-binding pockets of TfSGLr and CpSGL are similar in that the Sop 5 moiety in TfSGLr fits well in the substrate A novel ␤-1,2-glucanase from fungi pocket of CpSGL. There is a difference in the shape of the pocket between the two enzymes (Fig. 14B). The ␣-face side of the Glc moiety at subsite ϩ2 in TfSGLr is narrowed due to hydrophobic interaction with Tyr-311 and Trp-312, although this space in CpSGL is wide open and Phe-204 sticks out a little to the ␤-face side of the same Glc moiety in CpSGL. Such a difference might be related to the preferential Sop 2 release from ␤-1,2-glucans for TfSGL, unlike CpSGL that hydrolyzes ␤-1,2glucans in a random manner to release Sop 2-5 .
We compared the reaction mechanisms of TfSGL and CpSGL, although the catalytic mechanism of CpSGL has not been determined. The catalytic residues in TfSGLr and candidate residues in CpSGL are shown in Figs. 15 and 16A. The carboxyl groups of a general acid in TfSGLr (Glu-262) and Glu-211 in CpSGL are well-superimposed, although the positions of their main chains are different. Both residues are highly conserved in their respective homologs. In addition, these two acidic residues were found to be conserved on structure-based pairwise alignment of TfSGLr and CpSGL (Fig. 15). It has been reported that the activity of the E211Q mutant toward ␤-1,2glucan is drastically decreased as compared with the WT CpSGL (the relative activity is 0.15%) (25). These facts imply that CpSGL homologs may have the same general acid as TfSGL. On the contrary, a catalytic base in TfSGL (Asp-446) is substituted with Ile-399 in CpSGL. Ile-399 is highly conserved as a hydrophobic residue (Ile, Leu, and Val) in GH144 homologs. Furthermore, candidate residues for catalysis in CpSGL (Asp-135, Asp-139, and Glu-142) are located at the positions where proton transfer only via water molecules from a nucleophilic water is impossible according to the superimposed Sop 5 moiety in TfSGLr. These observations suggest that the general bases are obviously different between TfSGL and CpSGL despite the undetermined reaction mechanism of CpSGL.
The overall positions of substrate recognition residues in TfSGL and CpSGL are quite different (Fig. 16, B and C), although both enzymes have comparable hydrogen bonds at subsites Ϫ3, ϩ1, and ϩ2. At subsite Ϫ1, where no ligands are observed in CpSGL, there are three potential hydrogen bonds with His-119, Glu-142, and Glu-211, whereas TfSGLr forms five hydrogen bonds. In addition, there is no potential hydrogen bond at subsite Ϫ2 in CpSGL. Meanwhile, there are only two similar parts as follows: the Glc moieties are stacked on the aromatic residues at subsite Ϫ3 (Fig. 14B), and the carboxylate groups of Glu-262 in TfSGL and Glu-211 in CpSGL are located at almost the same position.
Overall, TfSGLr has the same substrate specificity, and a similar overall structure and shape of the catalytic pocket as CpSGL. However, TfSGLr is clearly different from CpSGL in primary sequence, the positions of the base catalysts, and most substrate recognition residues. Therefore, TfSGL and its homologs should be classified into a novel family, GH162.

Speculated physiological roles of TfSGL homologs
Almost all of the TfSGL homologs are distributed in Eukaryotes, especially in the Ascomycota, Basidiomycota, and Mycetozoa. Among the Ascomycota possessing TfSGL homologs, there are many species related to the rhizosphere such as Talaromyces verruculosus (62). Fusarium oxysporum is a plantpathogenic fungus, and Oidiodendron maius is known as an endophyte on azaleas. Metarhizium species and Beauveria bassiana are insect pathogens but reside in the rhizosphere, where they supply nitrogen to plants from insects. Considering that cyclic ␤-1,2-glucans are produced in the rhizosphere by plant symbionts such as Rhizobium, TfSGL homologs in such species might be involved in residence in the rhizosphere by metabolizing cyclic ␤-1,2-glucans.
Besides them, many pathogenic, symbiotic, and predatory species are known. In the Ascomycota, Cordyceps confragosa, Torrubiella hemipterigena, Aschersonia aleyrodis, and Tolypocladium ophioglossoides are fungi parasitic on specific insects and mushrooms. Pochonia chlamydosporia and Purpureocillium lilacinum are nematophagous fungi. Mycetozoa species prey on bacteria. In addition, there are also Ciliophora species such as a predator of bacteria (Paramecium tetraurelia) and a symbiotic Eukaryote found with a coral (Symbiodinium microadriaticum). It has been reported that some Paramecium spp. such as Paramecium bursaria form symbiotic relationships with symbiotic Chlorella (63). Considering the fact that cyclic ␤-1,2-glucans are reported to reduce the immune responses of hosts (9), TfSGL homologs might be related to interactions with other organisms. A novel ␤-1,2-glucanase from fungi Talaromyces cellulolyticus, a species closely-related to T. funiculosus, possesses a TfSGL homolog (GAM34680.1; the amino acid sequence identity with TfSGL is 99%). There are BGL (GH1 and GAM34681.1) and sugar transporter (GAM34682.1) genes in the vicinity of the gene encoding the TfSGL homolog. The BGL and transporter genes are also highly conserved in the vicinity of many TfSGL gene homologs. In addition, BGLs from T. reesei, an Ascomycota fungus, Trire2_120749 and Trire2_22197 (the amino acid sequence identities with the BGL from T. cellulolyticus are 49 and 69%, respectively) are highly up-regulated in the presence of Sop 2 (64). Therefore, TfSGL and its homologs may play a role in the metabolism of ␤-1,2-glucans with the cooperative action of the BGLs and transporters, although the degrading and binding activities of these putative BGLs and transporters have not been examined.
This study makes a significant contribution to expansion of GH families and the variety of reaction mechanisms for GH enzymes. Furthermore, this study will help us to clarify the molecular evolution of SGLs from Prokaryotes and Eukaryotes. The unique distribution of the new GH162 family will lead to exploration of the physiological roles of GH162 enzymes.

Materials
T. funiculosus (NBRC100958) and P. pastoris (KM71H) were purchased from the National Institute of Technology and Evaluation (NITE, Tokyo, Japan) and Thermo Fisher Scientific, respectively. Linear ␤-1,2-glucans with the average DP of 25 and 77 (unless otherwise noted, the average DP of ␤-1,2-glucans is 77) and Sop n s with DP of 2-11 were prepared using LiSOGP and CpSGL, as described previously (20,25). Cyclic ␤-1,2-glucan with DP of 17-24 was kindly donated by Dr. M. Hisamatsu of Mie University (65). rSop n s with DP of 6 -11 were prepared by modifying Sop n s at the reducing end with NaBH 4 treatment as described by Shimizu et al. (26). Approximately 10% Sop n solutions were added to 1 M NaBH 4 of 20 l, followed by incubation at room temperature for 5 min or more. After 3 M acetate (15 l) had been added to them, the rSop n s were precipitated with isopropyl alcohol (1 ml). Each pellet was dissolved in a small amount of water. Then, the samples were precipitated again with isopropyl alcohol and dried up. Laminarin and carboxymethyl (CM)-cellulose were purchased from Sigma. CM-pachyman, CM-curdlan, lichenan, ␤-glucan from barley, tamarind xyloglucan, glucomannan, arabinogalactan, A novel ␤-1,2-glucanase from fungi arabinan, and polygalacturonic acid were purchased from Megazyme (Wicklow, Ireland). Pustulan was purchased from Calbiochem.

Purification of SGL from T. funiculosus
T. funiculosus was grown in 2 liters of medium (comprising 0.5% ␤-1,2-glucan with the average DP of 25), 2 g/liter KH 2  and 100 g/ml ampicillin; adjusted to pH 6.0) (28) at 30°C in a shaking incubator (150 rpm) for 3 days. This culture was filtrated using a gauze cloth and a glass filter and concentrated using a Vivaflow 200 (10,000 molecular weight cutoff) (Sartorius, Gottingen, Germany). An ammonium sulfate solution containing 50 mM acetate-Na buffer (pH 5.5) was added to obtain 40% saturated ammonium sulfate concentration. The supernatant filtrated using a Ministart syringe filter (Sartorius) was loaded onto a HiTrap TM butyl HP column (5 ml; GE Healthcare, Buckinghamshire, UK) equilibrated with 50 mM acetate-Na buffer (pH 5.5) and 40% saturated ammonium sulfate (buffer A). After the unbound proteins had been washed with buffer A, a target protein was eluted using a linear gradient of ammonium sulfate (40 -0% saturated concentration). All fractions containing SGL activity were collected, buffered with 50 mM MOPS-NaOH buffer (pH 7.0), and then concentrated using Amicon Ultra 30,000 molecular weight cutoff (Merck Millipore, Darmstadt, Germany). Next, the sample was loaded onto a RESOURCE TM Q column (1 ml; GE Healthcare) equilibrated with 50 mM MOPS-NaOH buffer (pH 7.0). After unbound proteins had been washed out with the same buffer, the target enzyme was eluted with a linear gradient of 0 -1 M NaCl in 50 mM MOPS-NaOH buffer (pH 7.0). Finally, the enzyme solution concentrated with Amicon Ultra 30,000 molecular weight cutoff to 2 mg/ml (500 l) was loaded onto a Superdex TM 200GL column (24 ml; GE Healthcare) equilibrated with 50 mM acetate-Na buffer (pH 5.5) containing 150 mM NaCl, and then the target enzyme was eluted with the same buffer. All purification steps were carried out using an AKTA prime plus chromatography system (GE Healthcare).

Assay of SGL and BGL activity in the purification steps
To detect SGL activity in fractionated samples during the purification steps, the reaction was performed in 100 mM acetate-Na buffer (pH 5.5) containing 0.5% ␤-1,2-glucan and 50 mM D(ϩ)-glucono-1,5-lactone (GDL, Wako, Osaka, Japan) (pH 5.5) as an inhibitor of BGL and 35% (v/v) each enzyme solution at 30°C for an hour and was stopped at 100°C for 5 min. The Sop n s released from ␤-1,2-glucan were detected by TLC. In the case of the BGL assay, the reactions were performed using 5 mM p-nitrophenyl-␤-D-glucopyranoside (pNP-Glc, Wako) without GDL. After 9 volumes of 0.5 M Na 2 CO 3 had been added to a reaction mixture, the mixture was added to a 96-well microtiter plate (Sigma), and then the absorbance at 405 nm of the sample was measured using a Spectramax 190 (Molecular Devices, CA) (unless otherwise noted, the absorbance of samples was measured with this instrument). One unit was defined as the amount of the enzyme required to release 1 mol of pNP from pNP-Glc in a minute. The amount of pNP was calculated using the molar extinction coefficient of pNP (ϭ 18,200 M Ϫ1 cm Ϫ1 ).
The p-hydroxybenzoic acid hydrazide (PAHBAH) method was used for evaluation of ␤-1,2-glucandegrading activity A novel ␤-1,2-glucanase from fungi (66). It should be noted that the evaluated activity includes BGL activity, because this method measures the reducing power of the products. Reactions were performed with 5% (v/v) of the enzyme solution in 100 mM acetate-Na buffer (pH 5.5) containing 0.5% ␤-1,2-glucan at 30°C for an hour, and then 4 volumes of a PAHBAH solution (comprising 10 mg/ml PAHBAH, 0.1 M HCl, and 0.4 M NaOH) were added to the reaction mixtures. After heat treatment at 100°C for 5 min, the absorbance at 405 nm was measured. Glc was used as a standard for reducing sugar, and 1 unit was defined as the amount of the enzyme required to release 1 mol of the Glc equivalent reducing power of Sop n s in a minute.

SDS-PAGE and glycosylated protein analyses
The fractionated samples obtained during the purification were separated on 10% SDS-polyacrylamide gels (67). The gels were stained with 2D-silver stain reagent (Cosmo Bio, Tokyo, Japan). In the case of the recombinant enzymes, Coomassie Brilliant Blue R-250 was used for staining. Precision Plus Protein TM unstained standards (Bio-Rad) were used as molecular weight markers. Glycosylated protein analysis of the purified native TfSGL was performed as described below. N-Linked glycans in the native TfSGL were completely removed using glycopeptidase F (Takara Bio, Shiga, Japan) under denaturation conditions. After the samples had been loaded for SDS-PAGE, glycans and proteins in the gels were stained with a Pro-Q Emerald 300 gel stain kit (Thermo Fisher Scientific) and SYPRO Ruby protein gel stain (Thermo Fisher Scientific), respectively, according to the manufacturer's instructions. Avidin (Wako) and glucose oxidase (Oriental Yeast, Tokyo, Japan) were used as glycoprotein markers.

TLC analysis
The native TfSGL and TfSGLr were incubated in 100 mM acetate-Na buffer (pH 4.0) containing 0.2% each substrate (Sop n s with DP of 3-7, rSop n s with DP of 6 -8, cyclic ␤-1,2glucan, and linear ␤-1,2-glucan) at 30°C. After heat treatment at 100°C for 5 min, the reaction mixtures (0.5 l) were spotted onto TLC Silica Gel 60 F 254 (Merck Millipore) plates. The plates were developed with 75% acetonitrile. As for hydrolysates of Sop n s and rSop n s, the plates were developed twice with a solution (acetonitrile/acetic acid/isopropyl alcohol/deionized water ϭ 17:4:4:3). Then, the plates were soaked in a 5% (w/v) sulfuric acid/ethanol solution and heated in an oven until the spots were visualized clearly.

Size-exclusion chromatography
SEC analysis was performed as described above. Ovalbumin (44 kDa), conalbumin (75 kDa), aldolase (158 kDa), ferritin (440 kDa), and thyroglobulin (669 kDa) (GE Healthcare) were used as molecular weight markers. Blue dextran 2000 (2,000 kDa) was used to determine the void volume of the column. The molecular weight of TfSGL was calculated using Equation 1, where K av is the gel-phase distribution coefficient; V e is the volume required to elute each protein; V o is the volume required to elute blue dextran 2000; and V t is the bed volume of the column.

Identification of an SGL gene from T. funiculosus
To identify the amino acid sequence encoding an SGL gene, analysis of N-terminal and internal amino acid sequences of the purified enzyme was entrusted to Genostaff (Tokyo, Japan) and APRO Life Science Institute (Tokushima, Japan), respectively. Degenerate primers were designed based on the results of these sequence analysis (Table S4).
gDNA and total RNA were isolated from T. funiculosus cells grown in 100 ml of medium containing a 0.5% linear ␤-1,2glucan with the average DP of 25 as a sole carbon source at 30°C for 3 days (28). The collected cells were suspended in deionized water, frozen with liquid nitrogen, and then ground into powder. For gDNA preparation, a powdered sample was resuspended in the extraction buffer comprising 100 mM NaCl, 1% SDS, 2% Triton X-100, 2 units of RNase A, and 10 mM Tris-EDTA buffer (10 mM Tris-HCl and 1 mM EDTA) (pH 8.0). The solution was mixed with phenol/chloroform, and then ethanol was added to the water phase to precipitate the gDNA. After the precipitate had been dissolved in deionized water, 2 units of RNase A was added to the solution. This solution was incubated at 37°C for 1.5 h and then mixed with phenol/chloroform. The gDNA was precipitated by the addition of isopropyl alcohol to the water phase. The pellet was washed with 70% ethanol and then dissolved in the Tris-EDTA buffer after drying up. The total RNA was extracted from the disrupted cells using ISOGEN (Wako), and the cDNA was synthesized from the total RNA using ReverTra Ace quantitative PCR RT Master Mix with a gDNA remover (Toyobo, Osaka, Japan), according to the manufacturer's instructions.
PCR was carried out using TaKaRa Ex TaqHS (Takara Bio). Basically, the reactions were performed under the following conditions: one cycle of denaturation at 98°C for 2 min, 35 cycles of 98°C for 10 s, 50°C for 30 s, and 72°C for 0.5-1.5 min, and then one cycle of 72°C for 3.5 min. The scheme for cloning of the whole TfSGL gene was shown in Fig. 3. To amplify the gDNA region between internal peptides 1 and 2, primary PCR was performed using an F1-R1 primer pair (annealing temperature of 45°C) and gDNA as a template, and the amplified product was purified using a High Pure PCR product purification kit (Roche Applied Science, Basel, Switzerland). Then, nested PCR was performed using an F2-R2 primer pair and the purified primary PCR product as a template. This PCR product was purified and sequenced. R3 and R4 primers were designed as specific primers based on the sequence. A degenerated primer F3 was designed based on the N-terminal peptide sequence. Primary and nested PCR were performed as described above using an F3-R3 primer pair and an F3-R4 primer pair (extension time, 2 min) in that order, and then the amplified product was purified and sequenced. Finally, F4 and R5 primers were designed based on the analyzed nucleotide sequence to amplify the 5Ј-and 3Ј-regions of the gene encoding TfSGL. After gDNA had been digested with EcoRI or HindIII (Takara Bio), selfligation of the purified digests was performed using Ligation High, Version 2 (Toyobo). Inverse PCR was performed using an F4 -R5 primer pair (annealing temperature of 60°C, extension A novel ␤-1,2-glucanase from fungi time of 3 min and 30 cycles) and the self-ligated sample as a template. The PCR product was purified and sequenced. Then, partial regions in the TfSGL gene were amplified using specific primer pairs (F5-R5, F6 -R6, and F7-R7) and cDNA as a template to cover the whole ORF, and then the PCR product was sequenced to obtain the whole TfSGL gene sequence. The ORF, its transcription start site, and the polyadenylation signal sequence in the sequenced gDNA region were predicted with Softberry (http://www.softberry.com/) 3 (68).

Cloning, expression, and purification of TfSGLr using P. pastoris
The TfSGL gene without the region encoding the residues at the N terminus (1-21 amino acids) was amplified by PCR using KOD-Plus (Toyobo) as a DNA polymerase according to the manufacturer's instructions. First, PCR was performed using an F8 -R8 primer pair and the cDNA pool of T. funiculosus as a template. Second, PCR was performed to add a His 6 tag to the N terminus of the TfSGL gene using an F9 -R9 primer pair. The amplified product digested with EcoRI and NotI (Takara Bio) was ligated into the pPICZ␣B vector (Thermo Fisher Scientific) digested with the same restriction enzymes to fuse the secretion signal peptide for P. pastoris at the N terminus of the His 6 tag. The plasmid produced in E. coli XL1-blue was linearized with PmeI (New England Biolabs). The plasmid sample was purified with a phenol/chloroform/isoamyl alcohol solution (Sigma) and by ethanol precipitation. The purified plasmid was transformed into P. pastoris in the presence of 0.5 mg/ml Zeocin. TfSGLr was expressed according to the manufacturer's instructions for an EasySelect TM Pichia expression kit (Thermo Fisher Scientific). The transformant was harvested using 2 liters of buffered minimal glycerol medium at 30°C for about 18 h (until A 600 ϭ 2-6), and then the cells were suspended in 200 ml of buffered minimal methanol medium. Methanol (0.5%) was added to the suspension to induce TfSGLr, and enzyme induction was performed at 30°C for 4 -6 days. MOPS-NaOH buffer containing NaCl was added to the culture filtrate until the concentrations reached 500 mM NaCl and 50 mM MOPS-NaOH buffer (pH 7.0), and then the mixture was filtrated with a 0.45-m filter (Sartorius). The sample was loaded onto a His-Trap FF crude column (5 ml; GE Healthcare) equilibrated with 50 mM MOPS-NaOH buffer (pH 7.0) containing 500 mM NaCl (buffer B). After unbound proteins had been washed out using the same buffer containing 10 mM imidazole, TfSGLr was eluted using a linear gradient of imidazole concentration (10 -300 mM) in buffer B. The enzyme solution was buffered with 5 mM acetate-Na buffer (pH 4.0) using Amicon Ultra 30,000 molecular weight cutoff. The absorbance at 280 nm of the sample was measured using a spectrophotometer V-560 (Jasco, Tokyo, Japan), and the concentration of the enzyme was calculated using the theoretical molecular mass of TfSGLr (56,554 Da) and a molar extinction coefficient of 123,020 M Ϫ1 ⅐cm Ϫ1 (72).

General properties
To determine the optimum pH, TfSGLr (3.2 g/ml) was incubated in 100 mM various buffers containing 0.2% ␤-1,2glucan at 30°C for 15 min and then heated at 100°C for 5 min to stop the reaction. The reducing power of Sop n s released from the ␤-1,2-glucan was measured by the PAHBAH method. The optimum temperature for TfSGLr was determined after the enzyme had been reacted in 100 mM acetate-Na buffer (pH 4.0) at each temperature (0 -70°C). To examine the pH stability for TfSGLr, the purified enzyme (6.4 g/ml) was incubated in 20 mM various buffers at 30°C for an hour, and then the reaction was carried out in 100 mM acetate-Na buffer (pH 4.0) containing 0.2% ␤-1,2-glucan at 30°C for 15 min. To determine the temperature stability, TfSGLr (6.4 g/ml) was incubated in 100 mM acetate-Na buffer (pH 4.0) at each temperature (0 -70°C) for an hour, and then the reaction was carried out under the same conditions as for pH stability.

Substrate specificity
TfSGLr (3.2 g/ml) was incubated in 100 mM acetate-Na buffer (pH 4.0) containing each substrate (0.2% ␤-glucan, 0.1% glucomannan, 0.2% tamarind xyloglucan, 0.2% arabinan, 0.1% polygalacturonic acid, 0.0125% lichenan, 0.0125% laminarin, 0.05% pustulan, 0.1% CMC, 0.05% CM-curdlan, 0.025% CMpachyman, 0.006% arabinogalactan, or 0.2% ␤-1,2-glucan) at 30°C for 15 min, and then the amount of oligosaccharides released from each polysaccharide was measured by the PAHBAH method. Glc was used as a standard. The GOPOD method was used for determination of activity releasing Sop 2 from the ␤-1,2-glucan, as described below. The reaction mixtures were incubated with BGL (1 mg/ml) from almonds (Oriental Yeast) at 50°C for 2 h in 100 mM acetate-Na buffer (pH 5.0) to hydrolyze Sop 2 specifically. Finally, the GOPOD solution (9 volumes of the mixture) was added to the mixture, followed by incubation at 45°C for 20 min. The amounts of Glc in the samples were calculated by measuring the absorbance at 510 nm of the samples. Glc was used as a standard. One unit was defined as the amount of the enzyme required to release 1 mol of Sop 2 in a minute.

Kinetic analysis
To determine the kinetic parameters of TfSGLr for ␤-1,2glucan and Sop 5 , Sop 2 released from these substrates was quantified. After TfSGLr (1.5 and 7.5 g/ml for hydrolysis of ␤-1,2glucan and Sop 5 , respectively) had been incubated at 30°C for A novel ␤-1,2-glucanase from fungi concentrations of Sop 5 (0.12-2.5 mM) or ␤-1,2-glucan (5.0 ϫ 10 Ϫ3 Ϫ0.13 mM (ϭ 6.3 ϫ 10 Ϫ2 Ϫ1.6 (mg/ml)), the reactions were stopped by heat treatment at 100°C for 5 min. Sop 2 -releasing activity was determined using the GOPOD method, as described in the previous section. The experimental data obtained on hydrolysis of ␤-1,2-glucan and Sop 5 were fitted to the Michaelis- where v is initial velocity; [E] is enzyme concentration; k cat is turnover number; [S] is ␤-1,2-glucan concentration; and K m is Michaelis constant, and the substrate inhibition Equation 3, where v is initial velocity; [E] is enzyme concentration; k cat is turnover number; [S] is Sop 5 concentration; K m is Michaelis constant; and K i is substrate inhibition constant. The Kaleida-Graph program Version 3.51 was used for their regression analysis.

Stereochemical analysis of the reaction products by 1 H NMR
To determine which reaction mechanism TfSGL follows, an anomer inverting or retaining mechanism, 1 H NMR analysis was performed on the reaction products released from ␤-1,2glucan by TfSGLr. TfSGLr (0.12 mg/ml) was incubated at room temperature in 5 mM acetate-Na buffer (pH 4.0) containing 1.5% ␤-1,2-glucan with the average DP of 25 and 90% D 2 O. 1 H NMR spectra of the sample were recorded using a Bruker Advance 600 spectrometer (Bruker BioSpin, Rheinstetten, Germany). The assignment of the signals of Sop 2-4 and linear ␤-1,2-glucan was based on the previous studies (19,25). Acetate in the buffer was used as an internal standard for calculation of the relative peak area. To use acetone as an external standard for calibration of chemical shifts, as performed in the previous study (25), a droplet of acetone was added to the reaction mixture after the reaction had finished. Because chemical shifts derived from the C1 proton at the reducing end of ␤-anomer products are completely masked by the chemical shift derived from H 2 O, the chemical shift of the C2 proton at the nonreducing end was used as the total signal of ␣and ␤-anomers of the reaction products. The C2 protons were observed at almost the same chemical shifts regardless of the anomer, as described by Abe et al. (25).

Polarimetric analysis of the reaction products
To confirm the reaction mechanism of TfSGL, the time course of the degree of optical rotation in the reaction mixture was monitored. TfSGLr (1 mg/ml) was incubated at room temperature in 20 mM acetate-Na buffer (pH 4.0) containing 2% ␤-1,2-glucan with the average DP of 25, and the degree of the optical rotation of the reaction mixture was measured using a Jasco p1010 polarimeter (Jasco). Several droplets of 35% aqueous ammonia were added at 235 s after the reaction start to enhance mutarotation of the anomers.

Analysis of the reaction products from 3-deoxy Sop n -derivatives with TfSGLr
3dSop 5-6 and 3ЈdSop 5 were synthesized using LiSOGP from 3 and 3Ј-deoxy-Sop 2 , which are Sop 2 derivatives having 3-deoxyglucose moieties at the reducing and nonreducing ends, respectively. These Sop 2 derivatives were prepared as described previously (73). The reaction for synthesis of 3dSop 5-6 or 3ЈdSop 5 was performed in 50 mM MOPS-NaOH buffer (pH 7.5) containing 10 or 25 mM ␣-D-glucose-1-phosphate, 5 mM of each Sop 2 derivative, and 2 mg/ml LiSOGP at 30°C for an hour or 8 days, respectively. The reactions were stopped at 80°C for 5 min, and the supernatants were collected after centrifugation. The supernatants were added to Amberlite TM MB-4 (Organo, Tokyo, Japan) to remove unreacted ␣-D-glucose-1-phosphate. These samples were concentrated to 10 l with a centrifugal evaporator EC-57CS (Sakuma, Tokyo, Japan). Then, 3dSop 2-6 and 3ЈdSop 2-5 were separated with a Shimadzu LC-9A highperformance liquid chromatograph equipped with an SCL-6B system controller and SIL-6B autoinjector (Shimadzu, Kyoto, Japan) using an Asahipak NH2P-50 4E column (Shodex, Tokyo, Japan) with distilled water and acetonitrile (ϭ 35:65 (v/v)) as the eluent at a flow rate of 1 ml/min and with detection with a Refractive Index Detector (RI-8010) (Tosoh Bioscience, Tokyo, Japan). The enzymatic reactions of TfSGLr for 3dSop 5 and 3ЈdSop 5 were performed in 50 mM acetate-Na buffer (pH 4.0) containing 2 mg/ml TfSGLr and each derivative at 30°C for 180 min. When using 3dSop 6 as a derivative, the concentration of TfSGLr and the reaction time were changed to 5 g/ml and 30 min, respectively. Although the amounts of 3dSop 5-6 and 3ЈdSop 5 were too small to be determined, their concentrations in the reaction mixtures are estimated to have been 0.5, 2.5, and 4.2 M, respectively, based on comparison with the peak area of standard Sop 5-6 . These samples were concentrated with a centrifugal evaporator, and the reducing ends of sugars were fluorescently labeled with 2-aminobenzamide. First, 2-aminobenzamide (136 mg) was added to sodium cyanoborohydride (35 mg), and then they were dissolved in methanol (350 l) and acetic acid (41 l) to prepare the fluorescent reagent. The samples (10 l) were mixed with 40 l of the fluorescent reagent, followed by incubation at 80°C for an hour. The mixtures were extracted with chloroform (200 l) and distilled water (200 l) after cooling to room temperature. The standard solutions (containing 1 mM Glc and Sop 2-6 , 3dSop 2 , 3dSop 3 , 3dSop 5 , 3dSop 6 , 3ЈSop 2 , or 3ЈdSop 5 ) were labeled in the same way. Finally, the supernatants were collected by centrifugation. The solutions (3dSop 5-6 and 3ЈdSop 5 ) were dried up and then the samples were dissolved in distilled water (20 l). The fluorescently-labeled samples of 10 l were loaded onto a TSKgel Amide-80 column (Tosoh Bioscience) and separated and analyzed with a Waters W600 HPLC solvent delivery system (Waters) with detection with a 2475 fluorescence detector (Waters) at 422 nm.

Crystallography
To remove N-glycans from the purified TfSGLr, the enzyme (5 mg) was treated with 50 units of endoglycosidase H (New England Biolabs) in 50 mM citric acid buffer (pH 5.5) at 20°C A novel ␤-1,2-glucanase from fungi overnight. The sample was purified using a HisTrap FF crude column as described above. The enzyme solution was buffered with 5 mM acetate-Na buffer (pH 4.0) using Amicon Ultra 30,000 molecular weight cutoff and then concentrated to 20 mg/ml. The initial screening for TfSGLr crystal was performed using Wizard classic 1 and 2 block (Rigaku, Tokyo, Japan). The initial crystal condition for TfSGLr was a mixture of TfSGLr (10 mg/ml, 1 l) and a solution (1 l, comprising 0.1 M sodium phosphate dibasic/citrate buffer (pH 4.2), 0.2 M NaCl, and 20% (w/v) PEG 8000). Iodinated derivatives of TfSGLr crystals were prepared by vaporizing iodine labeling (74) to determine the initial phase. The TfSGLr crystals for the labeling were prepared under a modified condition as follows. The enzyme solution (20 mg/ml, 1 l) was mixed with 3 volumes of a reservoir solution comprising 0.1 M phosphate-citrate buffer (pH 4.8), 0.1 M NaCl, and 20% (w/v) PEG 3350, followed by incubation at 25°C for 5 days. A drop of iodine/potassium iodide solution comprising 135 mg/ml iodine and 250 mg/ml potassium iodide (1.5 l) was placed beside a drop of TfSGLr. The drop was incubated at 25°C for 3 days to label the TfSGLr with iodine (74). The apo crystal for data collection was obtained at 25°C for 3 days by mixing TfSGLr (20 mg/ml, 0.8 l) and a reservoir solution (3 l, containing 0.1 M citrate-Na buffer (pH 4.1), 0.1 M NaCl, and 20% (w/v) PEG 2000).
The crystal for the Sop 2 complex was obtained at 25°C for 3 days by mixing TfSGLr (20 mg/ml, 1 l) and a reservoir solution (3 l, comprising 0.1 M phosphate-citrate buffer (pH 4.6), 0.1 M NaCl, and 20% (w/v) PEG 3350). The crystal for the Sop 7 complex was obtained at 25°C for 3 days by mixing the E262Q mutant (10 mg/ml, 1.6 l) and a reservoir solution (3 l, comprising 0.1 M citrate-Na buffer (pH 4.0), 0.1 M NaCl, and 25% (w/v) PEG 2000). These crystals were soaked in the reservoir solution supplemented with 25% (w/v) PEG 400 for cryoprotection and then soaked in a solution containing 50 mM Sop 2 or 5% (w/v) ␤-1,2-glucan with the average DP of 25, respectively. Each crystal was soaked in the reservoir solution supplemented with 25% (w/v) PEG 400 as a cryoprotectant and kept at 100 K in a nitrogen-gas stream during data collection. All X-ray diffraction data were collected on a beamline (BL-5A) at Photon Factory (Tsukuba, Japan).

MD simulation
Initial structures for MD simulation were taken from the crystal structures of TfSGL in the apo form (6IMU, chain A) and in complex with Sop 7 (6IMW, chain B). Because the apo crystal structure has two alternative conformations of His-316 (Fig.  9A), one adopts a different orientation from the Sop 7 complex and the other adopts the same orientation with the Sop 7 complex, the two initial structures (termed apo1 and apo2, respectively) were prepared. The N-and C-terminal residues were capped with acetyl and N-methyl groups, respectively. N-Acetylglucosamines (GlcNAc) derived from N-linked glycan were removed from the structures. Disulfide bonds were introduced between Cys-27 and Cys-516, between Cys-230 and Cys-251, and between Cys-345 and Cys-499. The mutation of catalytic acid residue (E262Q) in the Sop 7 complex was manually reverted to that of the WT. The protonation states of amino acid residues were determined basically based on PROPKA 3.1 (41) (assuming pH 4.0), except for a few residues of which the protonation states were determined based on visual inspection of the hydrogen bond network. In the apo2 and the Sop 7 complex, His-57 and His-316 were protonated only on the N⑀2 atoms, and the other histidine residues were protonated on both the N␦1 and N⑀2 atoms. In apo1, His-57 was protonated only on the N⑀2 atoms, and the other histidine residues were protonated on both the N␦1 and N⑀2 atoms. Glu-24, Glu-36, Glu-82, Glu-144, Glu-180, Glu-193, Glu-226, Glu-246, Glu-262, Glu-286, Glu-323, Glu-389, Glu-430, Glu-478, Glu-481, and Glu-503 of the apo and Sop 7 complex structures were protonated, whereas other glutamate residues were deprotonated. In both the apo structures, Asp-48, Asp-111, Asp-168, Asp-177, Asp-234, Asp-259, Asp-278, Asp-369, and Asp-405 were protonated, and the other aspartate residues were deprotonated. In the Sop 7 complex, Asp-48, Asp-111, Asp-177, Asp-234, Asp-259, Asp-328, and Asp-405 were protonated, and the other aspartate residues were deprotonated. All the arginine and lysine residues were protonated. The systems for MD simulation were constructed by immersing the initial models in cubic water boxes where the distance between protein atoms and the closest boundary was at least 10 Å. Chloride ions were added to the systems for neutralization. The LEaP module of Amber-Tools 18 (83) was used to construct the systems. Amber ff14SB (84) and GLYCAM06j (85) force-field parameters and the TIP3P model (86) were used for the protein, carbohydrate, and water, respectively. A novel ␤-1,2-glucanase from fungi MD simulations were performed using GROMACS 2018 (87). After energy minimization and equilibration, a production MD run was carried out for 100 ns. During the MD simulation, the temperature was controlled at 303 K using the velocity rescaling thermostat (88), and the pressure was controlled at 1.0 ϫ 10 5 pascal using the weak coupling method (89). Bond lengths involving the hydrogen atoms were constrained using the LINCS algorithm (90) to allow the use of a 2-fs time step. Electrostatic interaction was calculated using the particle mesh Ewald method (91). MD trajectories were recorded every 10 ps.
C␣ RMSD and RMSF values were calculated using the gmx_rmsd and gmx_rmsf modules of GROMACS, respectively. PCA was carried out for a combined ensemble composed of ensembles from the simulations of the apo2 and the Sop 7 complex using the method described previously (92). Conformational distributions of the apo2 and the Sop 7 complex were calculated by separately projecting the trajectories onto the principal axes. The gmx_rmsf, gmx_covar, and gmx_anaeig modules were used for PCA. Representative structures from the MD ensembles were determined based on cluster analysis using the gmx_cluster module with the C␣ RMSD cutoff of 0.8 Å.

Mutational analysis
The plasmids of TfSGLr mutants were constructed using a PrimeSTAR mutagenesis basal kit (Takara Bio) according to the manufacturer's instructions. PCRs were performed using appropriate primer pairs (Table S4) and the TfSGLr plasmid as a template. The transformation to P. pastoris and the expression and purification of TfSGLr mutants were performed in the same way as WT TfSGLr. The enzymatic reactions were performed basically in the same way as for determination of substrate specificity. The concentrations of mutants (0.001-2.6 mg/ml) and reaction time (0 -2.5 h) were changed depending on the mutants. Color development was performed using the GOPOD method as described under "Kinetic analysis." For several mutants, the MBTH method (colorimetric determination method for Sop n s) (39) was also used for determination of specific activity, as described below.

MBTH method
The WT TfSGLr and TfSGLr mutants (0.001-2.6 mg/ml) were incubated in 20 mM acetate-Na buffer (pH 4.0) containing 0.2% ␤-1,2-glucan at 30°C for 0 -2.5 h. After appropriate intervals, 15-l aliquots of the reaction mixtures were taken and heated at 100°C for 5 min. The samples were mixed with 15 l of 0.5 N NaOH and then an MBTH solution (15 l, comprising 3 mg/ml MBTH and 1 mg/ml 1,4-DTT). The mixtures were incubated at 80°C for 15 min. A solution (30 l, comprising 0.5% FeNH 4 (SO 4 ) 2 , 0.5% H 3 NSO 3 and 0.25 N HCl) was added to the mixtures, and then 75 l of distilled water was also added after cooling to room temperature. The absorbance at 620 nm was measured. Sop 2 (0.5-2.5 mM) was used as a standard for reducing sugar, and 1 unit was defined as the amount of the enzyme required to release 1 mol of the Sop 2 equivalent reducing power in a minute.