Unique active-site and subsite features in the arabinogalactan-degrading GH43 exo-β-1,3-galactanase from Phanerochaete chrysosporium

Arabinogalactan proteins (AGPs) are plant proteoglycans with functions in growth and development. However, these functions are largely unexplored, mainly because of the complexity of the sugar moieties. These carbohydrate sequences are generally analyzed with the aid of glycoside hydrolases. The exo-β-1,3-galactanase is a glycoside hydrolase from the basidiomycete Phanerochaete chrysosporium (Pc1,3Gal43A), which specifically cleaves AGPs. However, its structure is not known in relation to its mechanism bypassing side chains. In this study, we solved the apo and liganded structures of Pc1,3Gal43A, which reveal a glycoside hydrolase family 43 subfamily 24 (GH43_sub24) catalytic domain together with a carbohydrate-binding module family 35 (CBM35) binding domain. GH43_sub24 is known to lack the catalytic base Asp conserved among other GH43 subfamilies. Our structure in combination with kinetic analyses reveals that the tautomerized imidic acid group of Gln263 serves as the catalytic base residue instead. Pc1,3Gal43A has three subsites that continue from the bottom of the catalytic pocket to the solvent. Subsite −1 contains a space that can accommodate the C-6 methylol of Gal, enabling the enzyme to bypass the β-1,6–linked galactan side chains of AGPs. Furthermore, the galactan-binding domain in CBM35 has a different ligand interaction mechanism from other sugar-binding CBM35s, including those that bind galactomannan. Specifically, we noted a Gly → Trp substitution, which affects pyranose stacking, and an Asp → Asn substitution in the binding pocket, which recognizes β-linked rather than α-linked Gal residues. These findings should facilitate further structural analysis of AGPs and may also be helpful in engineering designer enzymes for efficient biomass utilization.

Arabinogalactan proteins (AGPs) are plant proteoglycans with functions in growth and development. However, these functions are largely unexplored, mainly because of the complexity of the sugar moieties. These carbohydrate sequences are generally analyzed with the aid of glycoside hydrolases. The exo-b-1,3-galactanase is a glycoside hydrolase from the basidiomycete Phanerochaete chrysosporium (Pc1,3Gal43A), which specifically cleaves AGPs. However, its structure is not known in relation to its mechanism bypassing side chains. In this study, we solved the apo and liganded structures of Pc1,3Gal43A, which reveal a glycoside hydrolase family 43 subfamily 24 (GH43_ sub24) catalytic domain together with a carbohydrate-binding module family 35 (CBM35) binding domain. GH43_sub24 is known to lack the catalytic base Asp conserved among other GH43 subfamilies. Our structure in combination with kinetic analyses reveals that the tautomerized imidic acid group of Gln 263 serves as the catalytic base residue instead. Pc1,3Gal43A has three subsites that continue from the bottom of the catalytic pocket to the solvent. Subsite 21 contains a space that can accommodate the C-6 methylol of Gal, enabling the enzyme to bypass the b-1,6-linked galactan side chains of AGPs. Furthermore, the galactan-binding domain in CBM35 has a different ligand interaction mechanism from other sugar-binding CBM35s, including those that bind galactomannan. Specifically, we noted a Gly fi Trp substitution, which affects pyranose stacking, and an Asp fi Asn substitution in the binding pocket, which recognizes b-linked rather than a-linked Gal residues. These findings should facilitate further structural analysis of AGPs and may also be helpful in engineering designer enzymes for efficient biomass utilization.
Arabinogalactan proteins (AGPs) are proteoglycans characteristically localized in the plasma membrane, cell wall, and intercellular layer of higher land plants (1), in which they play functional roles in growth and development (2). The carbohydrate moiety of AGPs is composed of a b-1,3-D-galactan main chain and b-1,6-D-galactan side chain, decorated with arabinose, fucose, and glucuronic acid residues (1, 2). The chain lengths and frequencies of side chains are different among plant species, organs, and stages of development (3), and the overall structures of the carbohydrate moieties of AGPs are not yet fully understood. Degradation of polysaccharides using specific enzymes is one approach to investigate their structures and roles. In this context, exo-b-1,3-galactanase (EC 3.2.1.145) specifically cleaves the nonreducing end b-1,3-linked galactosyl linkage of b-1,3galactans to release D-galactose (Gal). In particular, it releases b-1,6-galactooligosaccharides together with Gal from AGPs (4,5) and is therefore useful for structural analysis of AGPs.
Glycoside hydrolases are classified into families based on sequence similarity, whereas they are also divided into two major groups according to their catalytic mechanisms (i.e. inverting enzymes and retaining enzymes) (9,10). Inverting enzymes typically utilize two acidic residues that act as an acid and a base, respectively, and a hydroxyl group connected to anomeric carbon inverts from the glycosidic linkage after the reaction. GH43 enzymes are members of the inverting group and share conserved Glu and Asp as the catalytic acid and base, respectively (8), but GH43_sub24 enzymes lack the catalytic base Asp (8,11,12). In Ct1,3Gal43A (from Clostridium thermocellum), Glu 112 was thought to be the catalytic base (13), but in BT3683 (from Bacteroides thetaiotamicron), Glu 367 (corresponding to Glu 112 of Ct1,3Gal43A) was found not to act as a base but to be involved in recognition of the C-4 hydroxyl group of the nonreducing terminal Gal, and instead, Gln 577 is predicted to be the catalytic base in the form of an unusual tautomerized imidic acid (12). An example of GH lacking a catalytic base, endoglucanase V from P. chrysosporium (PcCel45A), is already known, and based on the mechanism proposed for this enzyme, it is possible that tautomerized Gln functions as a base in GH43_sub24 or that this Gln stabilizes nucleophilic water. PcCel45A lacks the catalytic base Asp that is conserved in other GH45 subfamilies (14), but it uses the tautomerized imidic acid of Asn as the base, as indicated by neutron crystallography (15). However, it is difficult to understand the situation in GH43_sub24, because no holo structure with a ligand at the catalytic center has yet been solved in this family. Moreover, no structure of eukaryotic GH43_sub24 has yet been reported.
In the present work, we solved the apo and liganded structures of Pc1,3Gal43A. Based on the results, we discuss the catalytic mechanism and the mode of ligand binding to CBM35 in the two-domain structure.

Results
Overall structure of Pc1,3Gal43A The crystal structure of the SeMet derivative of Pc1,3Gal43A was first determined by means of the multiwavelength anomalous dispersion method, and this was followed by structure determination of the ligand-free WT, the WT bound with Gal (WT_Gal), the E208Q mutant co-crystallized with b-1,3-galac-totriose (Gal3; E208Q_Gal3), and the E208A mutant co-crystallized with Gal3 (E208A_Gal3). Data collection statistics and structural refinement statistics are summarized in Tables 1 and  2, respectively. The recombinant Pc1,3Gal43A molecule is composed of a single polypeptide chain of 428 amino acids (Gln 21 -Tyr 448 ) with two extra amino acids, Glu 19 and Phe 20 , derived from the restriction enzyme cleavage site, which are disordered and thus were not observed. The protein is decorated with N-glycans because it was expressed in Pichia yeast. Up to three sugar chains are attached at Asn 79 , Asn 194 , and Asn 389 ; the attached chains vary in position and structure, and most contain one or two GlcNAc moieties.
Pc1,3Gal43A is composed of two domains, and ligands introduced by soaking or co-crystallization are located in a subsite of the catalytic domain or the binding site of CBM35 (Fig. 1). The N-terminal catalytic domain consists of a five-bladed b-propeller (Gln 21 -Gly 325 ), as in other GH clan-F enzymes, and the Cterminal domain (PcCBM35) takes a b-jellyroll fold (Thr 326 -Tyr 448 ) structure, as in previously reported CBM35s (16)(17)(18)(19)(20)(21)(22)(23)(24)(25). PcCBM35 contains one calcium ion near the end of the first b-strand on a different domain surface from the plane to which the ligand binds (Fig. 1). The structure of PcCBM35 is similar to those of other known CBM35s. The interface area is 686 Å 2 and includes many water molecules. The PDBePISA server (RRID:SCR015749) indicates that the enzyme forms a complex in the crystal, but this is an effect of crystallization, and the enzyme exists as a monomer in solution (data not shown).
Sugar-binding structure of the Pc1,3Gal43A catalytic domain The five-bladed b-propeller exhibits an almost spherical structure, and two central cavities are located at the ends of the pseudo-5-fold axis (Fig. 1). One of them contains the catalytic site and it is common in almost all GH43 enzymes. The catalytic site is located in the center of the five-bladed b-propeller, whose blades are formed by Gln 21 or Asn 22 -Leu 87 (I in Fig. 1), Ser 88 -Asp 155 (II in Fig. 1), Ser 156 -Gly 204 (III in Fig. 1), Ala 205 -Ser 247 (IV in Fig. 1), and Ala 248 -Asp 297 (V in Fig. 1). As shown in Fig. 2, the Gal3 molecule co-crystallized with the E208Q mutant occupies subsites 21, 11, and 12 of the catalytic site, from the nonreducing end to the reducing end. Gal 21 is located at the bottom of the catalytic cavity, and Gal 11 and Gal 12 extend linearly outwards. Gal 11 is half-buried in the cavity, whereas Gal 12 is exposed at the surface ( Fig. 2A).
Gal 21 adopts a 1 S 3 skew boat conformation and interacts with many residues via hydrogen bonds and hydrophobic interactions. As shown in Fig. 2 (B and C), the C-2 hydroxyl group of Gal 21 forms hydrogen bonds with NH 2 of Arg 103 and with OE1 of Gln 263 via water. In addition, this water molecule is bound with O of Gly 228 . The C-3 hydroxyl group of Gal 21 also forms a hydrogen bond with OE2 of Glu 57 via water. Glu 102 , Tyr 126 , Asp 158 , Gln 208 , Thr 226 , Trp 229 , and Gln 263 interact with Gal3 through hydrophobic interactions. Notably, Trp 229 supports the flat C3-C4-C5-C6 structure of Gal 21 , and Tyr 126 recognizes the C-6 methylol and C-4 hydroxyl groups, whereas Glu 102 recognizes the C-3 hydroxyl and C-4 hydroxyl groups. In Gal 11 (as shown in Fig. 2, B and C), the C-2 hydroxyl group forms a hydrogen bond with NE2 of Gln 208 and N of Gly 228 , whereas O5 forms a hydrogen bond with ND2 of Asn 180 , and C-6 hydroxyl group forms a hydrogen bond with OD1 of Asn 179 via  water. Tyr 126 , Arg 157 , Asn 180 , and Gln 208 interact hydrophobically with Gal. In Gal 12 (Fig. 2, B and C), the C-2 and C-4 hydroxyl groups form hydrogen bonds with OG1 of Thr 226 and ND2 of Asn 180 , respectively. In addition, Thr 226 interacts with Gal 12 through hydrophobic interaction. Furthermore, the glycosidic oxygen between Gal 11 and Gal 12 interacts with ND2 of Asn 180 through a hydrogen bond.
In the structure of WT_Gal, one Gal was found at subsite 21, taking a 4 C 1 chair conformation with a-anomeric conformation of the C-1 hydroxyl group (data not shown). The binding mode of Gal 21 is almost the same as that in E208Q_Gal3, but the C-1 hydroxyl group in the axial position forms hydrogen bonds with Gly 228 and Gln 263 . No Gal3 molecule was observed at the catalytic domain in the structure of the Gal3 co-crystallized E208A mutant.
To identify the catalytic residues, we examined the relative activity of WT and the six mutants toward b-1,3-galactobiose (Gal2) and Gal3. WT showed 5.58 6 0.35 and 11.15 6 0.39 units of activity (mmol of Gal/min/nmol of enzyme) toward Gal2 and Gal3, respectively, whereas the six mutants showed no detectable activity (Fig. S1), suggesting that these residues are all essential for the catalysis.
Sugar-binding structure of CBM35 in Pc1,3Gal43A Pc1,3Gal43A has one CBM35 domain at the C terminus. We previously reported that this enzyme has a CBM6-like domain (6), but it has been reclassified into the CBM35 family (7). The b-jellyroll fold domain is accompanied by a single calcium ionbinding site on a domain surface different from the surface to which the ligand at the end of the first b chain binds, and this corresponds to a conserved calcium ion-binding site in CBM35s. Some CBM35 modules bind another calcium ion at a site at the top of domain (16), but PcCBM35 lacks this second calcium ion-binding site (Fig. 1).
In E208A_Gal3, electron density of Gal3 was observed in the ligand-binding site of PcCBM35. As illustrated in Fig. 3A and Fig. S2, 2F o 2 F c omit maps showed that the binding mode of PcCBM35 with ligands is "exo-type," corresponding to type-C CBM (26). The asymmetric unit of E208A_Gal3 contained four Pc1,3Gal43A molecules, and each molecule binds to the nonreducing end of Gal3 (called Gal_site 1), as in other CBM35 modules. However, the middle Gal (Gal_site 2) and the reducing end Gal (Gal_site 3) are found in two main locations (Fig. 3), although residues involved in the interactions with the ligand in each molecule were mostly shared. The Gal_site 1 forms Glu57 Gln208 Glu57 Gln208 Figure 2. Gal3-binding mode at the catalytic site. A, surface structure of the catalytic center. Gal3 is represented as green (carbon) and red (oxygen) sticks. B, schematic diagram showing the interaction mode at the catalytic center. Black, red, and blue, carbon, oxygen, and nitrogen, respectively. Red lines indicate the hydrophobically interacting residues. This diagram was drawn with LigPlot1 (version 1.4.5). C, the 2F o 2 F c omit map is drawn as a blue mesh (0.8s). Residues are shown in white (carbon), red (oxygen), and blue (nitrogen). Gal3 is shown in green (carbon) and red (oxygen). Yellow dots, hydrogen bonds and/or hydrophobic interaction; red spheres, water molecules interacting with ligands or residues.
hydrogen bonds with Tyr 355 and Arg 388 and interacts hydrophobically with Leu 342 , Gly 354 , Tyr 438 , and Asp 441 . The Gal_site 2 interacts hydrophobically with Gly 383 and Asp 384 . The main ligand interaction in the Gal_site 3 involves Gly 409 and Gly 410 , but in addition to these residues, Asn 411 is also involved in ligand recognition in chain C (Fig. 4).

Ensemble refinement
To understand the fluctuation of ligands, ensemble refinements were performed with the refined models. This method produces ensemble models by employing a combination of Xray structure refinement and molecular dynamics. These models can simultaneously account for anisotropic and anharmonic distributions (27). Four different pTLS values of 0.6, 0.8, 0.9, and 1.0% were set for each model. Table 3 shows the statistical scores of the refinement with the most appropriate pTLS value for each model. Focused views of the catalytic site in the catalytic domain and the ligand-binding site of the CBM are shown in Figs. 5 and 6, respectively. Note that structures containing multiple molecules in the asymmetric unit (WT and E208A_Gal3) are found for all molecules in this paper.
In the catalytic site, the vibration levels of some residues were significantly different between the apo and holo forms. As shown in Fig. 5, Tyr 126 , Arg 157 , Asp 158 , Asn 179 , Asn 180 , Gln 181 , Trp 229 , and Gln 263 in the liganded structures (Fig. 5, B, C, F, G, J, and K) showed smaller vibrations than in the apo structures (Fig. 5, A, D, E, H, I, and L). These results indicate that sidechain fluctuations converge upon ligand binding. Comparison of the Gal-bond structure (i.e. WT_Gal; Fig. 5, B, F, and J) with the Gal3-bond structure (i.e. E208Q_Gal3; Fig. 5, C, G, and K) showed that the fluctuations of Glu(Gln) 208 , Asn 179 , and Thr 226 of E208Q_Gal3 were smaller than these in WT_Gal. Therefore, it can be inferred that these residues recognize the ligands at the plus subsites. The catalytic acid, Glu 208 , has two major conformations in WT and WT_Gal. These two conformations were also reported in the BT3683 structure (12). Thus, the movement of this residue appears to be important for catalysis. Gln 263 shows one conformation (Fig. 5, A-D) that is identical to the result of the ensemble refinement of Asn 92 , known as imidic acid in PcCel45A (Fig. S3). Glu 102 may distinguish nonreducing terminal Gal, because it interacts with the axial C-4 hydroxyl group of Gal 21 (12). The vibration degree of Glu 102 was different between WTs and mutants, so its conformation does not depend on the ligand localization, but reflects interaction with Glu 208 , which serves as a general acid. Asp 158 of WT and E208A_Gal3 show greater vibration than WT_Gal and E208Q_Gal3. The role of Asp 158 is thought to be a pK a modulator; therefore, its function and conformational stability might be related. Focusing on Fig. 5 (I-L), there were large differences in the fluctuation level of Trp 229 . E208Q_Gal3 (Fig. 5K) showed small movements of Trp 229 , but other structures showed much larger fluctuations (Fig. 5, I, J, and L). These results suggest that this Trp is normally flipped and forms a p-p interaction to anchor the ligand in the proper position upon arrival. A histogram of the dihedral angle is shown in Fig. S4.
As regards the ligand-binding site of the CBM, a comparison of each chain of the E208A_Gal3 asymmetric unit showed no significant difference in the vibration levels of each residue involved in ligand binding (Fig. 6). However, ensemble refinement revealed that Gal_site 1 and Gal_site 2 do not show huge fluctuations, whereas Gal_site 3 has many conformations. They include the same conformation of each Gal chain in X-ray crystallography. Interestingly, a spatial difference in fluctuations was observed between ligands bound to the catalytic site and to the ligand-binding site of CBM35 (Fig. 7). At the catalytic site, Gal 21 is anchored in the appropriate position, and Gal 12 appears to fluctuate in a planar fashion as it interacts with the surrounding residues. In the CBM, it was inferred that Gal_site 1 is fixed and Gal_site 3 is adsorbed at the appropriate location at the binding site while fluctuating in three dimensions.

Discussion
Most exo-b-1,3-galactanases belonging to GH43_sub24 possess CBMs that can be classified into CBM35 or CBM13 (8). In this study, we elucidated the structure of a b-1,3-galactanbinding module for the first time by solving the structure of a GH43_sub24 containing CBM35 and obtained the ligandbound structures of both the catalytic and sugar-binding domains of Pc1,3Gal43A. This is also the first study to reveal the structure of a eukaryotic exo-b-1,3-galactanase. This information will be useful to understand how the CBM35 module interacts with b-1,3-galactan in combination with the GH43_sub24 catalytic module.
Although catalytic residues such as Glu and Asp are conserved in GH43 as a catalytic acid and base, respectively, GH43_sub24 lacks such a base residue. Cartmell et al. (12) suggested that GH43_sub24 may use Gln in the base role via conversion to imidic acid or use an exogenous base or utilize the Grotthuss mechanism of catalysis (8,12). In this study, we measured the enzyme activity of six variants of the three residues speculated to be involved in the catalytic reaction. As shown in Fig. S1, production of Gal by the mutants was not detected by means of HPLC analysis, suggesting that all three residues are essential for the catalytic activity of Pc1,3Gal43A. Glu 102 , Glu 208 , and Gln 263 are speculated to serve in C-4 hydroxyl group recognition, as a catalytic acid, and as a catalytic base, respectively. These residues are well-conserved in GH43_sub24, as shown in Fig. S5.
Most of the residues that interact with ligands are conserved in these three enzymes. In subsite 21, all residues, Glu 57 , Glu 102 , Arg 103 , Tyr 126 , Asp 158 , Glu 208 , Trp 229 , and Gln 263 , of Pc1,3Gal43A are conserved, indicating that the binding mode at subsite 21 is fully conserved in GH43_sub24. Based on the results of ensemble refinement, Trp 229 showed huge fluctuation, especially in the apo structure (Fig. 5, I-L). Trp 541 of BT3683, which corresponds to Trp 229 of Pc1,3Gal43A, has a   (6,12,28), we considered that these residues are not related to the mechanism of side-chain accommodation. The bypass mechanism of Pc1,3Gal43A, which enables accommodation of the b-1,6-galactan side chain so that the b-1,3-galactan main chain can be cleaved, appears to depend on the orientation of the C-6 methylol group of Gal3 at each subsite. The C-6 methylol group of Gal 21 is exposed to the solvent, so that the side chain can be accommodated externally. The C-6 methylol groups of Gal 11 and Gal 12 are also exposed to the solvent, so that the enzyme should be able to cleave the b-1,3 linkage of continuously b-1,6-substituted galactan, and a similar situation has been reported for BT3683 (12). Moreover, there are spaces near the nonreducing terminal Gal in these enzymes (12). This enables the enzymes to degrade the main chain, even if the side chain contains multiple carbohydrates. Similarly, b-1,3-glucanases belonging to GH55 also bypass the b-1,6-glucan side chain and degrade b-1,3-glucan from the nonreducing end (29,30). Comparing the surface structure of the catalytic site of Pc1,3Gal43A with that of these GH55 exob-1,3-glucanases from P. chrysosporium (PcLam55A), we see that Pc1,3Gal43A has a small pocket-like space capable of accepting the C-6 side chain of Gal at subsite 21 (Fig. 9, A and  B). In addition, the C-6 methylol group of Gals, located at the positive subsites of Pc1,3Gal43A, are exposed to solvent in a similar manner to that reported for SacteLam55A, GH55 exob-1,3-glucanase from Streptomyces sp. SirexAA-E (Fig. 9, A Figure 5. Results of ensemble refinement at the catalytic site. Each model is divided into three parts for clarity. A (E and I), B (F and J), C (G and K), and D (H and L) show WT, WT_Gal, E208Q_Gal3, and E208A_Gal3, respectively. Although WT and E208A_Gal3 contained multiple molecules in an asymmetric unit, the results obtained with multiple molecules were considered as an ensemble of one molecule in the present study. Letters indicate the chain names. Atoms are indicated in the same colors as in Fig. 2. Gal3 of the structure of E208Q_Gal3 obtained by X-ray crystallography is arranged in each figure to maximize ease of comparison. and C). Structures capable of accepting nonreducing terminal Gal with b-1,6-linked Gal are conserved among GH43_sub24 of known structure (Fig. 8 and Fig. S5). In the nonbypassing GH3 Hypocrea jecorina b-glucosidase (HjCel3A), the C-6 hydroxyl group of nonreducing glucose is oriented toward the enzyme, introducing steric hindrance (Fig. 9D) (31). In other words, enzymes bypassing side chains have a space adjacent to C-6 of the nonreducing terminal sugar, and the positive subsites are particularly wide, allowing side chains of the substrate to be accommodated. In contrast, enzymes unable to bypass the side chain have no space next to the 21 subsite and have a narrow entrance to the catalytic site, so that they are unable to accommodate side chains (Fig. 9D).
Although the electron density of Gal3 was observed in the present study, Pc1,3Gal43A is proposed to have four subsites ranging from 21 to 13, based on biochemical experiments (6). As mentioned above, although Pc1,3Gal43A has a structure capable of accepting the C-6 side chain, its degradation activity toward b-1,3/1,6-galactan is only approximately one-fifth that of the linear b-1,3-galactan (6). This difference in reactivity may be due to the structure of the sugar. The b-1,3-galactan in solution has a right-handed triple helical structure with 6-8 Gal residues per turn (32,33), with the C-6 methylol group pointing outward to avoid collisions between the b-1,6bonded Gal side chains (32). However, as shown in Fig. S6, Gal3 bound to the catalytic site of Pc1,3Gal43A is anchored to the enzyme, so that the helix of the glycans differs from the usual state in solution. Therefore, the reason why the hydrolytic activity of Pc1,3Gal43A toward b-1,3/1,6-galactan is lower than that toward b-1,3-galactan may be interference between the b-1,6-Gal side chains as a result of changes in the helical state of the main chain.
Although the amino acid sequence similarity of CBM35s is not so high, important residues involved in ligand binding are well-conserved (17). The modules belonging to CBM35 can be divided into four clades according to the mode of ligand binding, and the diversity in ligand binding and in the calcium ioncoordinating residue account for the various ligand-binding specificities (17) (Fig. 10A). Moreover, the residues involved in ligand binding of PcCBM35 differ from those of CBM35, which binds to a-Gal of galactomannan. This CBM is one part of a protein predicted to be the b-xylosidase of C. thermocellum cellulosomal protein (Cte_2137; Fig. 10), which belongs to the same cluster as PcCBM35 (17). There are some differences between the residues interacting with a-Gal of Cte_2137 and those interacting with b-Gal of PcCBM35. For instance, the regions of Ala 352 -Tyr 355 and Tyr 438 -Asp 441 of PcCBM35 correspond to Val 39 -Gly 42 and Ser 136 -Asn 140 of Cte_2137, which are related to ligand specificity (Fig. 10). Especially, Asn 140 of Cte_2137 is not conserved but replaced by Asp 441 in PcCBM35 and is located at the bottom of the ligand-binding site.  Furthermore, Trp 108 of Cte_2137 plays a key role in sacking the pyranose ring (17), whereas in CBM35 of Pc1,3Gal43A, this Trp residue is replaced with Gly (Fig. 10B). In other words, although PcCBM35 and Cte_2137 are in the same cluster, the residues involved in ligand recognition are different, and this difference affects the discrimination between b-Gal and a-Gal and between galactan and galactomannan. It is still unclear how CBM35s acquire such variation of binding specificity within a similar binding architecture. However, a detailed understanding of the molecular mechanisms of polysaccharide recognition by CBM35 will be essential for efficient utilization of various types of biomass.
In conclusion, we have determined the crystal structure of the catalytic and binding domains of Pc1,3Gal43A with the aim of reaching a detailed understanding of the mechanism of substrate accommodation by side chain-bypassing galactanase. Pc1,3Gal43A uses Glu as the catalytic acid and Gln as the catalytic base and has a structure in which the side chain of the substrate does not interfere with the catalytic reaction, thus mak-ing it possible to degrade the b-1,3-galactan main chain of AGPs despite the presence of the b-1,6-galactan side chain. Thus, although polysaccharides have a variety of molecular decorations, it appears that the structures of the degrading enzymes enable them to recognize specific features of the substrate while accommodating the variations. The introduction of mutations in substrate recognition residues to create enzymes with altered substrate recognition properties is expected to be helpful in the structural analysis of AGP glycans and also for the preparation of useful oligosaccharides.

Experimental procedures
Expression of Pc1,3Gal43A and its mutants The E208Q, E208A, E102Q, E102A, Q263E, and Q263A mutants were constructed by inverse PCR using PrimeSTAR MAX (Takara, Tokyo, Japan). For crystallization, Pc1,3Gal43A WT, E208Q, and E208A from P. chrysosporium were expressed in P. pastoris and purified as reported previously (7). For a reactivity The degree of conservation of amino acid residues in the catalytic domain of GH43_sub24 was visualized using the ConSurf server (RRID:SCR002320), the query for BLAST was set to Pc1,3Gal43A, and the conservation degree was analyzed based on 150 amino acid sequences in the ConSurf server (47)(48)(49)(50)(51). The conservation degree is shown in graded color. Preservation degrees are shown in a gradient with cyan for the lowest degree of preservation and blue for the highest. B, catalytic domain comparison of Pc1,3Gal43A and two GH43_sub24 galactanases. Shown are the catalytic centers of E208Q_Gal3 of Pc1,3Gal43A (white, PDB code 7BYV), BT3683 (cyan, PDB code 6EUI), and Ct1,3Gal43A (pink, PDB code 3VSF). Red, blue, and yellow, oxygen atoms, nitrogen atoms, and sulfur atoms, respectively. Residue names are shown for Pc1,3Gal43A/BT3683/Ct1,3Gal43A. assay, WT and mutants were purified by using SkillPak TOYO-PEARL Phenyl-650M (c.v. = 5 ml, Tosoh, Tokyo, Japan) equilibrated with 20 mM sodium acetate buffer, pH 4.0, containing 1 M ammonium sulfate, and the enzymes were eluted with 20 mM sodium acetate buffer, pH 4.0, containing 0.7 M ammonium sulfate. SeMet-labeled Pc1,3Gal43A was expressed as reported previously (7).

Data collection and structure determination
Diffraction experiments for Pc1,3Gal43A crystals were conducted at the beamlines of the Photon Factory (PF) or Photon Factory Advanced Ring (PF-AR), High Energy Accelerator Research Organization, Tsukuba, Japan (Table 1). Diffraction data were collected using CCD detectors (Area Detector Systems Corp., Poway, CA, USA). Crystals were cryocooled in a nitrogen gas stream to 95 K. For data collection of the WT enzyme complexed with Gal3, Pc1,3Gal43A crystals were soaked in a drop containing 1% (w/v) Gal3 for 10 min before the diffraction experiment. The data were integrated and scaled using the programs DENZO and SCALEPACK in the HKL2000 program suite (34).
Crystal structure was determined by means of the multiwavelength anomalous dispersion method using a SeMet-labeled crystal (7). Initial phases were calculated using the SOLVE/ RESOLVE program (35) from five selenium atom positions. The resultant coordinates were subjected to the automodeling ARP/wARP program (36) in the CCP4 program suite (37), and manual model building and molecular refinement were performed using Coot (version 0.8.9, University of Oxford, Oxfordshire, UK) (38), REFMAC5 (version 7.0.063, Science and Technology Facilities Council, Swindon, UK) (39), phenix. refine (40), and phenix.ensemble_refinement (27,41,42) in the Phenix suite of programs (version 1.13-2998-000, Lawrence Berkeley National Laboratory, Berkeley, CA, USA) (43). The refinement statistics are summarized in Table 2. ; PDB code 3ZYZ). A, B, and C hydrolyze the main chain of b-1,3-galactan or b-1,3-glucan, bypassing b-1,6-branched side chains (6,29,30). D hydrolyzes four types of b-bonds, and it does not bypass side chains (31,52). The top panel shows the overall surface structure, and the bottom panel shows an enlarged view of the catalytic region. Orange dashed circles, space near the C-6 position of Gal or glucose at the nonreducing end. Figure 10. Sequence alignment of known CBM35s (A) and structure comparison between CBM35s of Pc1,3Gal43A (B) and Cte_2137 (C). A, taxon names are shown as scientific names, ligand specificity, and PDB code only for brevity. When the same enzyme contains two CBM35 domains, the taxon name is indicated with 1 on the N terminus and 2 on the C terminus. Gal, Glc, Man, Xyl, and Uronic, ligand specificities for Gal, glucose, mannose, xylose, and glucronic acid and/or galacturonic acid, respectively. Among these, 3ZM8, 6UEH, and 2BGO, which bind to Man, are type B CBMs, which show endo-type binding, whereas the other 14 are all type C CBMs, which show exo-type binding. The alignment was built by using MUSCLE on MEGAX: Molecular Evolutionary Genetics Analysis (53,54), and the figure was generated with ESPrint 3.0 (RRID:SCR006587) (55). Orange and green boxes represent ligand-binding and calcium ionbinding residues, respectively. B and C, ligand-binding residues of Pc1,3Gal43A (chain A of E208A_Gal3) and Cte_2137 (PDB code 2WZ8). Red and blue, oxygen and nitrogen, respectively. The green stick model represents Gal3. Crystal structure of fungal GH43 galactanase For the analyses of WT and ligand-bound structures, structural determination was conducted by the molecular replacement method with the MOLREP program (44) in the CCP4 program suite using the SeMet or ligand-free structure as the starting model. Bound sugars, water molecules, and crystallization agents were modeled into the observed electron density difference maps. Calcium ion was modeled based on the electron density map and the coordination distances. Three Nglycans were observed, and the identified sugars were modeled. The stereochemistry of the models was analyzed with LigPlot 1 (version 1.4.5) (45,46), and structural drawings were prepared using PyMOL (version 2.2.3, Schrödinger, LLC, New York).
Enzymatic activity assay of Pc1,3Gal43A and its mutants To evaluate the reactivity toward Gal2 and Gal3 of WT and each mutant, 20 nM enzyme was incubated with 0.263 or 0.266 mM galactooligosaccharides in 20 mM sodium acetate, pH 5.0, for 30 min at 30°C, respectively. The reaction was stopped by heating at 95°C for 5 min. The supernatant was separated with 75% (v/v) acetonitrile on a Shodex Asahipak NH2P-50 4E column (Showa Denko, Tokyo, Japan), and the amount of released Gal was determined by HPLC (LC-2000 series; Jasco, Tokyo, Japan) with a Corona charged aerosol detector (ESA Biosciences, now Thermo Fisher Scientific). One unit of enzyme activity was defined as the amount of enzyme that releases 1 mmol of Gal/1 min/1 nmol of enzyme under our experimental conditions.

Data availability
The structures presented in this paper have all been deposited in the Protein Data Bank (PDB) with the following codes: 7BYS, 7BYT, 7BYV, and 7BYX. All remaining data are contained within the article.