Structural basis of exo-β-mannanase activity in the GH2 family

The classical microbial strategy for depolymerization of β-mannan polysaccharides involves the synergistic action of at least two enzymes, endo-1,4-β-mannanases and β-mannosidases. In this work, we describe the first exo-β-mannanase from the GH2 family, isolated from Xanthomonas axonopodis pv. citri (XacMan2A), which can efficiently hydrolyze both manno-oligosaccharides and β-mannan into mannose. It represents a valuable process simplification in the microbial carbon uptake that could be of potential industrial interest. Biochemical assays revealed a progressive increase in the hydrolysis rates from mannobiose to mannohexaose, which distinguishes XacMan2A from the known GH2 β-mannosidases. Crystallographic analysis indicates that the active-site topology of XacMan2A underwent profound structural changes at the positive-subsite region, by the removal of the physical barrier canonically observed in GH2 β-mannosidases, generating a more open and accessible active site with additional productive positive subsites. Besides that, XacMan2A contains two residue substitutions in relation to typical GH2 β-mannosidases, Gly439 and Gly556, which alter the active site volume and are essential to its mode of action. Interestingly, the only other mechanistically characterized mannose-releasing exo-β-mannanase so far is from the GH5 family, and its mode of action was attributed to the emergence of a blocking loop at the negative-subsite region of a cleft-like active site, whereas in XacMan2A, the same activity can be explained by the removal of steric barriers at the positive-subsite region in an originally pocket-like active site. Therefore, the GH2 exo-β-mannanase represents a distinct molecular route to this rare activity, expanding our knowledge about functional convergence mechanisms in carbohydrate-active enzymes.

enzymes showed the ability to efficiently degrade both ␤-mannans and MOS, being named as exo-␤-mannanases. These enzymes represent an instrumental simplification of the enzymatic route to obtain mannose from ␤-mannans. The structural basis for this interesting mode of action was unveiled for the GH5 family (20); however, there is no information about this activity within the GH2 family, which is known to classically harbor ␤-mannosidases. The GH5 enzymes are typically known to act on polymeric substrates, and most of characterized members are endo-acting enzymes, usually exhibiting a cleft-like catalytic interface. On the other hand, GH2 enzymes are known to act on shorter oligosaccharides or on substitutions such as arabinosyl and galactosyl side chains and, in contrast to GH5 members, feature a pocket-like active site. Thus, although similar activities are observed in these families, the molecular adaptations involved in the functional specialization are probably diverse, considering the distinct structural architecture and active site topology. Knowledge of those structural changes in the catalytic interface might be instrumental to understand their specificities and to generate relevant data for rational enzyme engineering aimed at biotechnological applications.
In this context, we discovered and characterized the first exo-␤-mannanase from the GH2 family, revealing the molecular modifications in the active site that enabled this enzyme to efficiently cleave both ␤-mannans and MOS, producing mannose. According to crystallographic analyses, the conserved hydrophobic barrier in the pocket-like active site of GH2 man-nosidases is absent in XacMan2A, generating extra productive positive subsites.

XacMan2A is an exo-␤-mannanase from the GH2 family
The X. axonopodis pv. citri GH2 enzyme (GenBank TM code AAM37920.1) ORF comprises 2,691 bp encoding an 886-amino acid protein with a molecular mass of 99.48 kDa. The recombinant XacMan2A was successfully overexpressed in BL21(DE3) Escherichia coli cells, and the purification procedure yielded 2.5 mg of pure and homogenous enzyme per liter of culture, despite its large size and structural complexity. CD analysis indicated a proper folded conformation with a melting temperature of 47.6°C (Fig. 1A), which is similar to that observed in other CAZyme isolated from the same bacterium (21). XacMan2A thermal unfolding followed the canonical two-state model with a single transition (Fig. 1A), despite the five-domain organization inferred by sequence analysis and further confirmed by X-ray crystallography.
Activity assays with synthetic substrates revealed that XacMan2A is only active on p-nitrophenyl-␤-D-mannopyranoside (pNP-␤-Man) exhibiting optimum pH and temperature of 5.5 and 40°C, respectively, with a moderate thermotolerance (Fig. 1, B-D). Interestingly, increasing hydrolysis rates were observed from mannobiose (M2) to mannohexaose (M6), indicating the enzyme preference for longer manno-oligosaccharides (Table 1). In addition, XacMan2A was able to cleave mannose-based polysaccharides, showing high activity against nm by CD spectroscopy. The melting temperature (T m ) was calculated from the sigmoidal fit of the denaturation curve. Shown are effects of pH (B) and temperature (C) on relative activity of XacMan2A. D, thermal stability curve. Residual activities were estimated at 40°C in 40 mmol⅐liter Ϫ1 citratephosphate buffer, pH 5.5, using 40 mmol⅐liter Ϫ1 pNP-␤-Man as substrate. 100% relative activity corresponded to 35.3 Ϯ 1.2 units⅐mg Ϫ1 protein, estimated in the same conditions. Values shown represent means Ϯ S.D. (error bars) from triplicate assays carried out with three separate preparations of purified recombinant XacMan2A.
The cleavage patterns of MOS and polymeric substrates, analyzed by capillary zone electrophoresis, indicate an exo-acting mechanism with mannose as the main product (Fig. 2). The reduction of MOS (M5 and M6) with borohydride did not alter the activity, suggesting that the enzyme recognizes the nonreducing ends of the substrate.
Kinetic characterization with M2, M6, and ␤-mannan were performed (Fig. 3A); however, the saturation was not achieved with M2, supporting the adaptation of XacMan2A to cope with longer oligosaccharides and polysaccharides. On the other hand, XacMan2A efficiently cleaved both M6 and ␤-mannan substrates (Fig. 3, A and B). Remarkably, XacMan2A exhibited a k cat of ϳ12.5 s Ϫ1 on ␤-mannan (Fig. 3C), which is at least 3-fold higher than that observed for the GH5 exo-␤-mannanase (20), highlighting its improved capacity to act on polymeric substrates. Together, these analyses pointed out that XacMan2A is not a classic ␤-mannosidase but a genuine exo-␤-1,4-mannanase, revealing a new activity in the GH2 family.

The multi-modular structural architecture, biological assembly, and negative-subsite mapping
To understand the structural basis of XacMan2A mode of action, the crystallographic structure was determined in the native form and in complex with mannose ( Table 2). The XacMan2A structure consists of five well-defined domains (Fig.  4, A and B) with a (␤/␣) 8 catalytic core at the central position. This modular organization is similar to that seen in other GH2 enzymes, such as ␤-mannosidases, from Bacteroides thetaiotaomicron (BtMan2A) (22), Dictyoglomus thermoplilum (DtMan) (23), and Trichoderma harzianum (ThMan2A) (24), and ␤-galactosidases, from E. coli (25) and Arthrobacter sp. (26). XacMan2A domain I (Fig. 4, residues Ser 29 -Trp 234 ) comprises two antiparallel ␤-sheets and an ␣-helical motif. Domain II (residues Asp 235 -Arg 346 ) and domain IV (residues Ala 699 -Gln 800 ) are structurally similar and display a ␤-sandwich fold, which is commonly found in carbohydrate-binding modules (reviewed by Hashimoto (27)). The catalytic domain, domain III (residues Ser 347 -Phe 698 ), has a classical (␤/␣) 8 TIMbarrel fold, commonly observed in the clan of GH-A enzymes. Domain V, located at the C terminus, is the shortest one (residues Leu 801 -Glu 888 ) and is the most divergent compared with other structurally characterized GH2 enzymes.
In one of the crystalline forms of XacMan2A, a dimer was found in the asymmetric unit; however, structural and energetic analyses indicate that the dimer is not stable in solution, which was further confirmed by small-angle X-ray scattering (SAXS) analysis. In solution, XacMan2A exhibited a radius of gyration (R g ) of 51.5 Ϯ 0.1 Å, and the calculated low-resolution envelope was fully compatible with the crystallographic monomer (Fig. 5, A and B). Interestingly, the hydrodynamic behavior of XacMan2A differs from the GH2 ␤-mannosidase BtMan2A, which forms dimers in solution. The dimerization in BtMan2A involves extensive contacts between the domains V of each monomer (22) that are not present in the crystal packing of XacMan2A. In addition, the domain V of XacMan2A largely diverges, in structure and composition, from other GH2 enzymes (38% sequence identity with the same domain of BtMan2A), in particular at the dimerization interface ( Fig. 5C), which explains the inability of XacMan2A to dimerize.
The catalytic residues, initially inferred by structural comparisons, were further confirmed with the mutants E477A (acid-base) and E575A (nucleophile) that were shown to be inactive against 4-nitrophenyl-p-mannopyranoside and ␤-mannan polysaccharide. These residues are fully conserved in the family and adopt similar conformations in BtMan2A and ThMan2A, with ϳ3.5 Å of distance, which is compatible with the classical Koshland retaining mechanism (28,29) that is expected for members of the GH-A clan.
Aiming to obtain a complete description of enzymesubstrate interactions, crystals of inactive mutants were exhaustively prepared with MOS from M2 to M6 by soaking and co-crystallization experiments; however, no high quality crystallographic data were attained. Only mannose complexes were obtained for inactive and active forms of XacMan2A (Fig.  6). In all complexes, the monosaccharide was observed at the Ϫ1 subsite, which is surrounded by several tryptophan residues (Trp 410 , Trp 552 , and Trp 667 ) and the polar residues Asp 215 and Asn 476 (Fig. 6) that are fully conserved among the structurally characterized GH2 ␤-mannosidases.
In the WT-mannose complex, the O1 atom from mannose makes polar contacts with Glu 575 and Asn 476 , whereas the O5 atom interacts with Glu 477 (Fig. 6A). In the E575A mutant structure, the O2 atom interacts with Glu 477 and the O3 atom with Asn 476 . In addition, the residue Asp 215 makes contacts with the O3 and O4 atoms in the WT protein and with the O4 and O6 atoms in the E575A mutant. The mannose displacement in the E575A structure is probably due to Glu 575 sidechain absence, indicating a crucial role of the nucleophile residue in the proper positioning of substrate (Fig. 6B). The mannose molecule in the E477A mutant showed a very similar position compared with the WT complex, except for the lack of interaction between the Glu 477 side chain and the O5 atom from the sugar ring (Fig. 6C). Despite the differences and the displacement found in the E575A structure, in all mannose-

GH2 exo-␤-mannanase
complexed structures (WT and mutants), the mannopyranosid ring assumes the chair geometry 4 C 1 , which is the most stable configuration in solution (30, 31) (Fig. 6). It correlates with the mannose configuration bound to the Ϫ2 subsite of the ␤-mannanase from Cellvibrio japonicus (PDB code 2VX5) (32) and with the mannosyl residues from the mannotriose bound to Streptomyces sp. ␤-mannanase (PDB code 5JU9) (31). In GHs, the substrate saccharide unit that binds to the negative subsite generally adopts a boat-or skew-boat-type conformation before the cleavage (30), such as in retaining ␤-mannosidases/ mannanases that often use the 1 S 5 to B 2,5 to 0 S 2 (33) or the 1 C 4 to 3 H 4 to 3 S 1 (31) catalytic pathways. Because the XacMan2A (WT and mutants) complexes were done with the product, no such configurations in the monosaccharide were observed.

Structural adaptations in the positive-subsite region are determinant for the exo-␤-mannanase activity
Although the catalytic acidic residues and the Ϫ1 subsite are conserved in GH2 ␤-mannosidases and in XacMan2A (Fig. 7A), the positive-subsite region of the latter enzyme is largely divergent in terms of composition and conformation, which might explain its particular mode of action. An important feature of the typical pocket-like active site topology in GH2 enzymes is the presence of a physical barrier in the positive-subsite region formed mainly by the loops Ala 539 -Asp 571 in ThMan2A (24), and Pro 514 -Thr 527 in BtMan2A (22) (Fig. 7B). However, in XacMan2A, the corresponding region (residues Gly 534 -Thr 544 ) differs in length, composition, and 3D configuration, generating a continuous and more accessible active-site cleft (Fig. 7, B and C), suggesting the presence of more than two productive subsites, as expected based on MOS hydrolysis rates (Table 1). For instance, in ThMan2A (24), this loop is 32 residues long with a ␤-hairpin motif. In BtMan2A (22), the loop is 13 residues long and includes a tryptophan residue (Trp 519 ), precluding the continuity of a potential positive-subsite region. In XacMan2A, this loop is the shortest one among these GH2 members with only 10 residues, and none of them represents a barrier for the positive-subsite region.
To identify those additional positive subsites, we docked different MOS (M2, M4, and M6) in XacMan2A and performed molecular dynamics simulations to obtain the relative binding energies (Fig. 8, A and B). Calculated relative binding-free energies were nearly 2-fold higher from M2 to M4, and a similar trend was observed from M4 to M6, indicating a higher affinity of XacMan2A for longer substrates (Fig. 8B). These in silico results are in accordance with biochemical data and suggest the existence of potentially six subsites ranging from Ϫ1 to ϩ5 (Fig. 8A).
In addition to the profound changes in the positive-subsite region compared with other GH2 members, XacMan2A contains two residue substitutions that increase the active-site vol-

GH2 exo-␤-mannanase
umetric capacity, Gly 439 and Gly 556 , which correspond to Asp 451 and Tyr 575 in ThMan2A and to Cys 424 and Tyr 537 in BtMan2A, respectively (Fig. 9A). To understand the functional role of these two nonconserved residues, we constructed the mutants G556Y and G439C, mimicking the BtMan2A configuration. Biochemical assays showed that both mutations abolished the enzyme activity on ␤-mannan polysaccharide (Fig.  9B), revealing that the decrease in volumetric capacity impaired the polymeric substrate binding. In addition, an unexpected decrease in the activity was observed on pNP-␤-Man (Fig. 9C), indicating distinct structural determinants for substrate recognition in XacMan2A, since the residue Tyr 537 was considered essential for BtMan2A catalysis (22).
Together, our data demonstrate that the XacMan2A positive-subsite region clearly provides accessibility to the substrate with a cleft that can accommodate up to six sugar residues (Figs. 7C and 8A). These structural observations are in full agreement with the functional assays, supporting the novel mode of action of XacMan2A.

The (dis)similarities of the catalytic interface of GH2 and GH5 exo-␤-mannanases
XacMan2A has the same ability of the Cellvibrio mixtus GH5 exo-␤-mannanase (CmMan5A) (20) to release mannose from pNP-␤-Man and from the nonreducing end of ␤-mannans. Structural superposition of the two catalytic domains resulted in RMSD of 2.4 Å (computed for C ␣ atoms) with the acid-base and nucleophile residues located in equivalent positions and oriented in a similar manner (Fig. 10A). Other conserved residues are Arg 80 , Asn 214 , Trp 285 , and Trp 376 from the catalytic center of CmMan5A, which correspond to Arg 408 , Asn 476 , Trp 552 , and Trp 667 in XacMan2A (Fig. 10A). These residues participate in substrate binding and also in the maintenance of the hydrogen bond network that supports catalysis. However, the conserved WDW motif adjacent to the Ϫ1 subsite of GH2 mannosidases is absent in CmMan5A. Instead, CmMan5A possesses an extended loop (Trp 378 -Phe 412 ) that forms a "double" steric barrier at the negative-subsite region, preventing an endo-acting mode (20) (Fig. 10B).
Interestingly, most exo-activities on polymeric substrates and oligosaccharides reported so far for CAZymes seem to have a similar evolutionary strategy involving the transition from an endo-acting enzyme to one with exo-activity, such as in the GH8 (34), GH10 (21), and GH43 (35) families. In all three of these examples, the insertion of a loop blocked the positive subsites (Fig. 11), leading to an exo-cleavage pattern. The active site topology of these proteins could be classified as blocked clefts, because the negative-subsite region remains open, whereas the positive region is clogged (Fig. 11). The evolution of exo-␤-mannanase activity in GH26 and GH5 families also seems to follow a similar strategy but, in these cases, via the insertion of a blocking loop at the negative-subsite region (32,20). In all of these cases, the exo-activity probably originates from the emergence of steric barriers in one of the sides of a cleft-like active site. XacMan2A is the first member of a typical family of enzymes with catalytic preference for side-chain substitutions and short oligosaccharides that assumes an exo-activity against polymeric substrates via the removal of steric impediments in a pocket-like active site (Fig. 11).

Discussion
XacMan2A is an unusual GH2 member because of its ability to efficiently hydrolyze the polysaccharide ␤-mannan. The enzyme showed greater efficiency on ␤-mannan compared with other mannan-based polymeric substrates, indicating that glucose in the main-chain or galactose substitutions impairs its catalytic activity. Indeed, crystallographic analysis did not show any evidence that galactosyl substituents could be accommodate in the active site. Among the GH2 ␤-mannosidases characterized so far, the only one that showed some capacity to cleave ␤-mannan was BtMan2A (22). However, its k cat (0.49 s Ϫ1 ) on ␤-mannan is very low compared with that observed on pNP-␤-Man (128.15 s Ϫ1 ). In contrast, XacMan2A hydrolyzes pNP-␤-Man and ␤-mannan with catalytic rates in the same order of magnitude (v o /[E T ] of 24.5 s Ϫ1 to pNP-␤-Man and 8.8 s Ϫ1 to ␤-mannan; Table 1), characterizing a genuine exo-␤-mannanase. Moreover, increasing hydrolysis rates were

GH2 exo-␤-mannanase
observed from M2 to M6 for XacMan2A, which was not observed for BtMan2A (22). Quantitative MS kinetics analysis also showed a clear preference of XacMan2A for longer oligosaccharides. In addition, all MOS and ␤-mannan were hydrolyzed to mannose by XacMan2A, and the cleavage pattern is characteristic of an exo-acting enzyme. Together, these results demonstrate that XacMan2A acts as an exo-␤-mannanase.
The unique composition and conformation of the loop Gly 534 -Thr 544 in the positive-subsite region of XacMan2A resulted in a catalytic interface more open and accessible that is probably associated with the capacity to cleave longer substrates. Compared with structurally characterized GH2 ␤-mannosidases, this loop in XacMan2A is shortened and does not impose a physical barrier, allowing the emergence of a continuous active-site cleft with additional positive subsites. Structural analysis, along with biochemical data, indicates that XacMan2A has only one negative subsite (Ϫ1) and, potentially, at least five positive subsites, which fully meets the stereochemical conditions for its new mode of action. In the other GH2 enzymes, such as BtMan2A and ThMan2A, "blocking" loops are found in the same region of the active site, and they can be associated with the selection of smaller substrates. The mannose-releasing exo-␤-mannanase activity had been observed in a GH5 family member from C. mixtus (CmMan5A) (20). However, the structural determinants for this mode of action in CmMan5A were attributed to the emergence of a loop that blocked the negative-subsite region, whereas in XacMan2A, it is associated with the removal of a steric barrier at the positivesubsite region.
This rare ability of XacMan2A to saccharify ␤-mannans without relying on endo-␤-mannanases represents a simplified carbon-uptake strategy that would benefit the bacterial growth and development during plant infection. This activity is also of great potential in industrial applications involving ␤-mannan depolymerization, such as biofuel production and food and beverage processing. In addition, this novel molecular mechanism for exo-activity based on the redesign of a pocket-like active site expands our current understanding of the molecular strategies related to functional differentiation in CAZymes.

Molecular cloning, mutagenesis, and protein production
A fragment of 2,691 bp from the XAC3075 gene (NCBI accession code AAM37920), encoding for XacMan2A without the predicted signal peptide (28 N-terminal residues), was amplified from the genomic DNA of X. axonopodis pv. citri 306 using standard cloning methods. The amplified DNA was cloned into pGEM-T vector and then subcloned into the restriction enzyme NdeI/XhoI sites of the pET28a vector (Novagen, Madison, WI) with a hexahistidine tag at the N terminus. Site-directed mutagenesis of XacMan2A construct (E575A, E477A, G439C, and G556Y) were performed according to the QuikChange kit (Stratagene, La Jolla, CA). The recombinant

GH2 exo-␤-mannanase
proteins were expressed in E. coli BL21 (DE3) cells and purified by metal-affinity and size-exclusion chromatography. Cells were grown at 37°C in lysogeny broth medium containing kanamycin (50 mg⅐ml Ϫ1 ) to A 600 nm ϭ 0.6, followed by induction with 0.5 mmol⅐liter Ϫ1 isopropyl-thio-␤-D-galactopyranoside for 16 h at 20°C. After centrifugation, cells were resuspended in lysis buffer (20 mmol⅐liter Ϫ1 sodium phosphate (pH 7.4), 300 mmol⅐liter Ϫ1 NaCl, 5 mmol⅐liter Ϫ1 imidazole, 1 mmol⅐liter Ϫ1 phenylmethylsulfonyl fluoride, and 5 mmol⅐liter Ϫ1 benzamidine) and incubated on ice with lysozyme (1 mg⅐ml Ϫ1 ) for 30 min. Bacterial cells were disrupted by sonication, and the soluble fraction was purified by immobilized metal ion affinity chromatography using a 5-ml HiTrap Chelating HP column (GE Healthcare, Little Chalfont, UK), previously charged with Ni 2ϩ , coupled to an ÄKTA purifier (GE Healthcare). The proteins were eluted using a linear gradient (0 -500 mmol⅐liter Ϫ1 ) of imidazole at a flow rate of 1 ml⅐min Ϫ1 . The eluted fractions were analyzed by SDS-PAGE, and those containing pure proteins were pooled, concentrated by filtration, and submitted to size-exclusion chromatography. Size-exclusion chromatogra-

GH2 exo-␤-mannanase
absorbance at 280 nm using the molar extinction coefficient calculated from the amino acid composition (http://web.expasy.org/protparam/). Dynamic light scattering was performed in a Dynapro molecular sizing instrument (Wyatt Technology) to evaluate the homogeneity of the purified samples.

CD analysis
CD spectroscopy experiments were conducted on a Jasco J-810 CD spectrophotometer (JASCO, Oklahoma City, OK) equipped with a Peltier temperature control using 1-mm path quartz cuvettes. Spectra were acquired with 3.5 mol⅐liter Ϫ1 XacMan2A in 20 mmol⅐liter Ϫ1 sodium phosphate (pH 7.4) with 150 mmol⅐liter Ϫ1 NaCl at 20°C. The thermal denaturation of XacMan2A was followed by measurement of the ellipticity changes at 212 nm. For this experiment, a temperature ramp from 20 to 100°C with a heating rate of 1°C⅐min Ϫ1 was used. The reversibility of the temperature effect was checked by cooling the denatured sample to 20°C using the same parameters described above.

SAXS data collection and analysis
SAXS measurements were performed using a monochromatic X-ray beam ( ϭ 1.488 Å) from the D01A-SAXS2 beamline at the Brazilian Synchrotron Light Laboratory (LNLS, Campinas, Brazil). SAXS measurements for XacMan2A were performed at two different concentrations (2 and 4 mg⅐ml Ϫ1 ) in Figure 7. XacMan2A active-site topology differs from GH2 ␤-mannosidases. A, details of the conserved Ϫ1 subsite between XacMan2A (PDB code 6BYC), BtMan2A (PDB code 2JE8), and ThMan2A (PDB code 4CVU). Side chains of the active site residues are shown as sticks, and carbon atoms are displayed in green (XacMan2A), in cyan (BtMan2A), and in orange (ThMan2A). Negative and positive subsites are represented by minus and plus signs, respectively. B, representation of ThMan2A, BtMan2A, and XacMan2A active-site regions in yellow, highlighting the loops in ThMan2A (residues from Ala 539 to Asp 571 , in orange), BtMan2A (residues from Pro 514 to Thr 527 , in cyan), and XacMan2a (residues from Gly 534 to Thr 544 ) that contribute to determine the active-site topology (shown in yellow). C, surface representation of ThMan2A (orange), BtMan2A (cyan), and XacMan2A (green), highlighting the active-site vicinity, with some delimiting residues represented as sticks. The second Trp residue from the conserved WDW motif was used as reference for the active-site length measurements. 20 mmol⅐liter Ϫ1 sodium phosphate buffer, pH 7.4, with 150 mmol⅐liter Ϫ1 NaCl. The samples were centrifuged for 20 min at 20,000 ϫ g and 4°C to remove potential aggregates before each measurement. The integration of SAXS patterns was performed using Fit2D (36). The package GNOM (37) was used to evaluate the pair-distance distribution functions p(r). Molecular envelopes were calculated from the experimental SAXS data using the program DAMMIN (38). Averaged models were generated from several runs using DAMAVER suite programs (39). The theoretical scattering curve was calculated from the crystallographic model and compared with the experimental SAXS curves using the program CRYSOL (40). The crystallographic structures were fitted into the corresponding SAXS molecular envelopes using the program SUPCOMB (41).

GH2 exo-␤-mannanase Protein crystallization
Crystals of native XacMan2A and mutants were grown by the hanging-drop vapor-diffusion method at 18°C in drops containing 0.5-1.0 l of protein solution at 7.2 mg⅐ml Ϫ1 and 0.5 l of the crystallization condition equilibrated against 200 l of crystallization condition in 48-well plates. The WT enzyme was crystallized in the following conditions: 0.1 mol⅐liter Ϫ1 ammonium acetate, 0.1 mol⅐liter Ϫ1 BisTris, pH 5.5, and 17% (w/v) PEG 10,000; and 0.1 mol⅐liter Ϫ1 ammonium acetate, 0.1 mol⅐liter Ϫ1 BisTris, pH 5.5, and 18% (w/v) PEG 10,000. A complex of XacMan2A with D-mannose (Megazyme, County Wicklow, Ireland) was prepared with the crystal grown in the second crystallization condition, by soaking the crystal into the mother solution containing 100 mmol⅐liter Ϫ1 D-mannose for 4 h. Before data collection, the crystal of native enzyme was soaked in cryoprotectant containing the mother solution with 20% (v/v) PEG 400 and then directly flash-cooled in a nitrogen gas stream at 100 K. Mutant E575A crystal was obtained in 0.1 mol⅐liter Ϫ1 sodium acetate, pH 5.0, 0.2 mol⅐liter Ϫ1 magnesium chloride, and 20% (w/v) PEG 600. Mutant E477A crystal was obtained in 0.1 mol⅐liter Ϫ1 ammonium acetate, 0.1 mol⅐liter Ϫ1 BisTris, pH 5.5, and 18% (w/v) PEG 10,000. Mutant complexes were prepared as described for the WT protein.

X-ray data collection, structure determination, and refinement
The X-ray data were collected on the MX2 beamline at the LNLS (Campinas, São Paulo, Brazil) using a 1.459 Å wavelength X-ray beam and a PILATUS2M detector (Dectris, Baden-Dattwil, Switzerland). Data were scaled and reduced using XDS (42). The structures were solved by molecular replacement methods using the crystalline structure of the GH2 ␤-mannosidase from B. thetaiotaomicron (PDB code 2JE8 (22); sequence identity of 39%) as template and MOLREP (43) in the CCP4 suite of programs. The models were further built with the Auto-Build wizard (44) from the PHENIX package, yielding nearly refined structures without internal gaps in the chain. The structures were refined alternating cycles of TLS and restrained refinement using REFMAC (45) or PHENIX Refine (46) and manually inspected using the program COOT (47). TLS groups were generated by the TLSMD server (48). The refined structures were evaluated using the program MolProbity (49). Data collection, processing, and refinement statistics are summarized in Table 2. Graphic representations of structures were generated in PyMOL (50). Structure superposition RMSDs were calculated with PDBeFOLD (51).
The structure factors and atomic coordinates of XacMan2A in native and mannose complex forms were deposited in the Figure 11. Molecular mechanisms adopted by exo-enzymes derived from classical endo-or exo-GH members. The GH8 exo-oligoxylanase (PDB code 1WU4 (34)) presents an emergent loop (red cartoon) formed by the residues Ser 317 -Leu 323 that blocks the positive-subsite region, resulting in a blocked cleft (represented in the scheme). A similar mechanism is adopted by the GH10 exo-oligoxylanase (PDB code 4PMV (21)) (residues Val 267 -Pro 303 ) and by the GH43 exo-arabinanase (PDB code 4KCB (35)) (residues Arg 203 -Asn 231 ). In the GH26 (PDB code 2VX5 (32)) and GH5 (PDB code 1UUQ (20)) exo-mannanases, a steric barrier imposed by the emergent loops (Arg 201 -Asn 231 and Trp 378 -Phe 412 , respectively) at the negative-subsite regions confers the exo mode of action and also results in a blocked cleft. In the XacMan2A (PDB code 6BYC), a nonblocking loop (green cartoon) (residues Gly 534 -Thr 544 ) at the positive-subsite region disrupts the classical pocket found in GH2 enzymes and gives rise to extra positive subsites. All emergent blocking loops are represented in red cartoons. Active-site cavities are highlighted in yellow. Negative (NS) and positive subsites (PS) are represented by a minus and plus sign, respectively.

GH2 exo-␤-mannanase
Protein Data Bank under the accession codes 6BYC and 6BYE, respectively. The crystallographic data of nucleophile (E575A) and acid-base (E477A) mutants were deposited under the accession codes 6BYG and 6BYI, respectively.

Molecular docking and in silico binding energy evaluation
The complexes between the target protein and MOS (M2, M4, and M6) were built using local small molecule docking onto the active site with AutoDock Vina (52). Based on the top hit, the binary complexes were prepared for molecular dynamics simulations using explicit solvent in YASARA (53) under standard temperature and pressure parameters. The binding energies were calculated throughout the trajectories according to the YAMBER3 force field (54).

Biochemical assays
Aryl glycosidase activity was determined at 40°C in 40 mmol⅐liter Ϫ1 citrate-phosphate buffer, pH 5.5, containing p-nitrophenyl-␤-D-glucopyranoside (pNP-Glu), p-nitrophenyl-␤-D-galactopyranoside (pNP-Gal), p-nitrophenyl-␤-D-xylopyranoside (pNP-Xyl), p-nitrophenyl-␣-L-arabinofuranoside (pNP-AraF), p-nitrophenyl-␣-D-mannopyranoside (pNP-␣-Man), or pNP-␤-Man (Sigma-Aldrich) as substrate in a final volume of 0.3 ml. The reactions were initiated by the addition of the enzyme and interrupted after convenient time intervals by adding 0.3 ml of saturated sodium tetraborate solution. The released p-nitrophenol was measured at 400 nm, using an Infinite 200 PRO microplate reader (TECAN Group Ltd., Männedorf, Switzerland). The hydrolysis of mannose oligosaccharides, ivory nut mannan, carob galactomannan (low viscosity), konjac glucomannan (low viscosity), sugar beet arabinan, beechwood xylan, carboxymethyl cellulose 4 M (CMC), and barley ␤-glucan (low viscosity) (all from Megazyme) were evaluated under the same conditions by estimating the reducing sugar released, according to the 3,5-dinitrosalicylic acid method (55). The optimum temperature for enzymatic activity was determined in the range from 10 to 85°C in 40 mmol⅐liter Ϫ1 citrate-phosphate buffer, pH 5.5. The optimum pH was evaluated in McIlvaine buffer (56), ranging from 3.0 to 8.0 at 40°C. Thermal stability was evaluated by incubating the enzyme in McIlvaine buffer, pH 5.5, for up to 32 h at 40°C. The residual activity was estimated at 40°C in 40 mmol⅐liter Ϫ1 citrate-phosphate buffer, pH 5.5, as described above. Reduced MOS samples were previously treated with borohydride and purified with a PD MiniTrap TM G-10 column (GE Healthcare) before enzymatic assays. One unit of enzyme was defined as the amount of enzyme that releases 1 mol of product per min. Specific activity was defined as units⅐mg Ϫ1 protein (units⅐mg Ϫ1 ). Kinetic parameters were calculated using nonlinear regression with Origin version 8.1 software (OriginLab, Northampton, MA).

Enzymatic activity monitored by MS
Hydrolysis reactions were carried at 40°C with 5 l of MOS (M6 and M2), 3 l of McIlvaine buffer at pH 5.5, 1 l of water, and 1 l of a solution of XacMan2A enzyme (final concentration 15 g/ml). The reactions were initiated by the addition of the enzyme and interrupted after 5 min by adding 40 l of methanol. Kinetic analysis were performed based on Ge et al.
(57) on a Synapt high-definition mass spectrometer (Waters Corp., Milford, CT) at V mode and ESI(ϩ) with a spray voltage maintained at 3.0 kV and heated to 130°C in the source. The quenched samples were injected in scan mode (m/z 150 -1,100) with direct infusion at a flow rate of 20 l/min. A calibration curve was made to determine the concentrations of the products of the enzymatic reaction. Kinetic parameters were calculated using nonlinear regression with Origin version 8.1 software (OriginLab).

Capillary zone electrophoresis
The products of enzymatic hydrolysis of mannose oligosaccharides with varying degrees of polymerization from 2 to 6 and ␤-mannan polysaccharides were derivatized with 8-aminopyrene-1,3,6-trisulfonic acid (Sigma-Aldrich) by reductive amination as described previously (58,59). Capillary zone electrophoresis was performed on a P/ACE MQD capillary electrophoresis system with a laser-induced fluorescence detection system (Beckman Coulter, Brea, CA). An uncoated fused-silica capillary of internal diameter 75 m and length 20 cm was used for separation of labeled oligosaccharides. The capillary was conditioned with 40 mmol⅐liter Ϫ1 potassium phosphate (pH 2.5), and samples were injected by application of 0.5 p.s.i. for 5 s. Electrophoretic conditions were 20 kV/70 -100 mA with reverse polarity at a controlled temperature of 25°C. 8-Aminopyrene-1,3,6-trisulfonic acid-labeled saccharides were excited at 488 nm, and emission was collected through a 520-nm band pass filter. The electrophoretic behavior of degradation products was compared with mannose oligosaccharide standards (purchased from Sigma-Aldrich and Megazyme).