Structural insights into the catalytic mechanism of a novel glycoside hydrolase family 113 β-1,4-mannanase from Amphibacillus xylanus

β-1,4-Mannanase degrades β-1,4-mannan polymers into manno-oligosaccharides with a low degree of polymerization. To date, only one glycoside hydrolase (GH) family 113 β-1,4-mannanase, from Alicyclobacillus acidocaldarius (AaManA), has been structurally characterized, and no complex structure of enzyme–manno-oligosaccharides from this family has been reported. Here, crystal structures of a GH family 113 β-1,4-mannanase from Amphibacillus xylanus (AxMan113A) and its complexes with mannobiose, mannotriose, mannopentaose, and mannahexaose were solved. AxMan113A had higher affinity for −1 and +1 mannoses, which explains why the enzyme can hydrolyze mannobiose. At least six subsites (−4 to +2) exist in the groove, but mannose units preferentially occupied subsites −4 to −1 because of steric hindrance formed by Lys-238 and Trp-239. Based on the structural information and bioinformatics, rational design was implemented to enhance hydrolysis activity. Enzyme activity of AxMan113A mutants V139C, N237W, K238A, and W239Y was improved by 93.7, 63.4, 112.9, and 36.4%, respectively, compared with the WT. In addition, previously unreported surface-binding sites were observed. Site-directed mutagenesis studies and kinetic data indicated that key residues near the surface sites play important roles in substrate binding and recognition. These first GH family 113 β-1,4-mannanase–manno-oligosaccharide complex structures may be useful in further studying the catalytic mechanism of GH family 113 members, and provide novel insight into protein engineering of GHs to improve their hydrolysis activity.


Overall structure of AxMan113A
The structure of AxMan113A was solved and the crystallographic statistics are summarized in Table 2. The orthorhombic space group of AxMan113A was P2 1 2 1 2 1 with two monomers in the asymmetric unit ( Fig. 2A). The monomer with approximate dimensions of 41 ϫ 31 ϫ 56 Å contained residues from Met-1 to Arg-309. AxMan113A displayed the typical (␤/␣) 8 -barrel-fold of GH family 113 members, containing an internal core of 8 ␤-strands encircled by 8 ␣-helices. The catalytic proton donor and nucleophile (Glu-143 and Glu-223) were determined by structural alignment to another GH family 113 ␤-1,4-mannanase (AaManA, PDB entry 3CIV), located in the middle of the narrow groove (Fig. 2B).

Crystal structures of AxMan113A-E223A complexes
The structures of AxMan113A-E223A complexes with M2, M3, M5, and M6 were obtained by co-crystallization method. In the AxMan113A-E223A-M2 structure, one M2 molecule was embedded in the narrow groove and occupied subsites Ϫ2 and Ϫ1 in chain B. In chain A, however, three mannose units were found in subsites Ϫ2, Ϫ1, and ϩ2, indicating the existence of two binding modes in AxMan113A-E223A-M2, occupying subsites Ϫ2 to Ϫ1 and ϩ1 to ϩ2. The lack of electron density for mannose moieties might have been due to the following: (i) mannoses in subsites ϩ1 and ϩ2 were very flexible; (ii) binding domains in subsites ϩ1 and ϩ2 were too narrow to completely bind one M2 molecule. In subsites Ϫ1 and Ϫ2, M2 residues formed 12 direct hydrogen bonds to Trp-11, Asn-89, Arg-96, Glu-143, Tyr-195, Asn-237, and Tyr-291. Three water-mediated hydrogen bonds were formed between Asn-237 and the mannose residues. Specifically, the amino groups of the side chains of Trp-11, Asn-89, and Arg-96 were directly hydrogen-

Structural analysis of GH family 113 ␤-1,4-mannanase
AxMan113A showed that most residues were in identical positions, except for Trp-239 and Asp-240. The side chain of Trp-239 was deflected in the former, forming a hydrophobic platform to bind to the mannose residues in subsites Ϫ2 and ϩ2. The distance between the side chain of Asp-240 and ϩ1 mannose was 3.3 to 2.7 Å, suggesting that Asp-240 was attracted during the recognition and binding with substrates. Interestingly, some additional electron-density maps representing M2 molecules were found on the surface region of the AxMan113A-E223A-M2 complex structure, which is far away from the catalytic cleft. In chain A, one M2 molecule was found in the middle of two ␣-helixes. Another M2 molecule was observed on the extended loop of chain B, and neither of these mannoses were structurally symmetrical ( Fig. 4). A crystal structure of AxMan113A-E223A in complex with M3 (AxMan113A-E223A-M3) was solved at 1.68-Å resolution ( Table 3). Only one M2 molecule was found to occupy to the catalytic cleft of AxMan113A (at subsites Ϫ2 to Ϫ1), suggesting that mannose occupies subsites Ϫ2 to Ϫ1 preferentially during the process of substrate binding (Fig. 5A). The complex structure of AxMan113A-E223A-M5 was determined at 2.38-Å resolution. Four mannose units (M4) linked by ␤-1,4-glycosidic bonds could be built into the electron-density map (at subsites Ϫ4 to Ϫ1). Mannose residues in subsites Ϫ3 and Ϫ4 extended to the terminal and lacked interactions with any surrounding key amino acids, indicating that AxMan113A can only weakly bind Ϫ3 and Ϫ4 mannoses (Fig. 5B). Superposition of AxMan113A-E223A-M5 and AxMan113A-E223A-M2 showed that the Ϫ1 and Ϫ2 mannose residues were basically in almost the same position, and all adopted the chair conformation 4 C 1 . At least six binding sites existed in the groove, occupying subsites Ϫ4 to ϩ2. In addition, the complex structure of AxMan113A-E223A-M6 was solved. Similarly, one M3 molecule was well-defined in the electrondensity maps, occupying the catalytic cleft of AxMan113A (subsites Ϫ1 to Ϫ3) (Fig. 5C).

Discussion
Because of the various biochemical properties, ␤-1,4-mannanases are widely used in many industrial applications. In paper/pulp industries, ␤-1,4-mannanases are able to cleave the mannan portion of pulps to increase the brightness (2, 7). In the field of clinical applications, ␤-1,4-mannanases can be used to hydrolyze nondigestible oligosaccharides and polysaccharides like prebiotics and dietary fibers, producing short chain fatty acids, which can protect against inflammatory bowel disease, ulcerative colitis, and Crohn's disease (1,15). ␤-1,4-Mannanases was also used to produce manno-oligosaccharides, which can effectively regulate the immune system and reduce serum cholesterol (16). In the field of functional foods, high viscosity of guar gum can be hydrolyzed by ␤-1,4-mannanases to yield partially hydrolyzed guar gum, which offers health benefits and nutrient digestion (17). Due to their growing industrial potential, many of ␤-1,4-mannanases have been studied. GH family 113 only had one structureresolved ␤-1,4-mannanase (AaManA) from A. acidocaldarius Tc-12-31 (10). AaManA has been reported to share a typical (␤/␣) 8 -barrel folding motif and displays a two-step retaining cata- Structural analysis of GH family 113 ␤-1,4-mannanase lytic mechanism, which is consistent with ␤-1,4-mannanases in GH families 5 and 26. The crystal structure of ␤-1,4-mannanasetargeted inhibitors (AaManA-ManIFG-M2) was solved after AaManA (18). However, lack of enzyme-manno-oligosaccharide complex structural information gives only limited comprehensive clarification of the catalytic mechanism of GH family 113 members. In this study, we cloned a novel GH family 113 ␤-1,4-mannanase gene from A. xylanus (AxMan113A) and obtained four manno-oligosaccharide complex crystals to further elucidate the catalytic mechanism of GH family 113 members.
The catalytic mechanism of GH family 113 ␤-1,4-mannanases can be described as a two-step retaining catalytic mechanism (25). The catalytic residues are Glu-143 and Glu-223 in AxMan113A, as confirmed by site-directed mutagenesis and structural alignment to AaManA. Previous studies have shown that GH family 113 members hydrolyze mannooligosaccharides with DP Ͼ 3 and exhibit considerable transglycosylation activity (10,19). However, the smallest substrate of AxMan113A was M2 (Fig. S4B). In addition, AxMan113A had no transglycosylation activity. Thus, the catalytic mechanism of AxMan113A differs from that of other GH family 113 members. AxMan113A could hydrolyze the reduced manno-oligosaccharides with DP Ն 4, and the hydrolysis efficiency toward M5r was higher than that of M4r (Fig. S5). It indicated that the reducing terminal substitution by mannitol could influence the recognition of mannose residue in the catalytic groove. Moreover, the hydrolysis of AxMan113A was diminished gradually with the substitutions approached to the catalytic groove. That might be one major reason that this enzyme cannot hydrolyze pNPM.
To verify the unique catalytic mechanism of AxMan113A, crystal structures of AxMan113A and its complexes with M2, M3, M5, and M6 were solved (Table 3). Clearly, mannose residues bound to at least 6 subsites in the groove (Ϫ4 to ϩ2) (Fig.  S6B). In the structure of AxMan113A-E223A-M6, one M3 molecule was occupied in Ϫ3 to Ϫ1 subsites, and another 3 substrate-binding sites (ϩ1 to ϩ3) has been proved by hydrolyzing M5r (Fig. S5). Mannose moieties of M5r would preferen- PDB code 5YLH 5YLK 5Z4T 5YLI 5YLL a R merge ϭ ͚ hkl ͚ i ͉I i (hkl) Ϫ ͗I(hkl)͉͘/͚ hkl ͚ i I i (hkl), where I i (hkl) is the ith observation of reflection hkl and ͗I(hkl)͘ is the weighted average intensity for all observations i of reflection hkl. b R work/free ϭ ͚ hkl ͉͉F obs ͉ Ϫ k͉F calc ͉͉/͚ hkl ͉F obs ͉; R work is the R value for the reflections used in the refinement, whereas R free is the R value for 5% of the reflections selected randomly and not included in the refinement.

Structural analysis of GH family 113 ␤-1,4-mannanase
tially bind in subsites Ϫ2 to ϩ2. Mannitol was occupied in subsite ϩ3 (24, 28). Thus, it could be deduced that another M3 molecule was occupied in ϩ1 to ϩ3 subsites. A bulk hydrogenbond network was formed around the subsite Ϫ1, which firmly fixed the position of the Ϫ1 mannose. Three residues (Arg-96, Tyr-195, and Trp-273) of GH family 113 ␤-1,4-mannanases were conserved (Fig. 1). Arg-96 not only hydrogen bonded with the Ϫ1 mannose, but also formed one hydrogen bond and salt bridge with the side chain of Asp-143 at a distance of 2.9 Å, stabilizing the latter's orientation. Similarly, the presence of Tyr-195 (2.7 Å away from Glu-223) was vital to orienting catalytic residue Glu-223 (Fig. S7A). This is consistent with studies of chitosanases OU01 and N174 (30,31). Trp-273 in AxMan113A provided a hydrophobic sugar-binding platform with the ϩ1 sugar ring. The importance of three residues for substrate-binding was corroborated by site-directed mutagenesis. Mutants R96A, Y195A, and W273A lost most of this activity (Fig. S7C). Compared with AaManA-ManIFG-M2, one key water molecular in AxMan113A-E223A-M2 was found above Ϫ1 subsite, forming three hydrogen bonds with Ϫ1 mannose, better fixing the sugar's position. This suggests that the binding of Ϫ1 mannose in AxMan113A is stronger than in AaManA. Moreover, in subsite ϩ1, the conformation of three residues (Trp-95, Lys-176, and Asp-240) in AxMan113A changed a great deal compare with AaManA-ManIFG-M2. Trp-95 flipped up ϳ30°, which brought the aromatic nucleus close to subsite ϩ1, strengthening the interaction with ϩ1 sugar. The side chains of Lys-176 and Asp-240 (His-183 and Arg-248 in AaManA, respectively) extended to subsite ϩ1, forming more hydrogen bonds with the mannose (Fig. S7B). Specific activity of W95A, K176A, and D240A was decreased to 7.1, 9.6, and 10.2% of the WT, respectively (Fig. S7C). AxMan113A exhibited higher affinity for Ϫ1 and ϩ1 mannoses, which might be an important reason why the enzyme can hydrolyze M2. As already noted, unlike other members of GH family 113, AxMan113A had no transglycosylation activity. In the GHs with retaining catalytic mechanism, some amino acids near subsite ϩ2 play an important role in stabilizing the transition state of transglycosylation (32). Compared with AaManA-ManIFG-M2, the conformation of mannose in the subsite ϩ2 of AxMan113A changed significantly. This suggested that the ϩ2 sugar is very flexible, making its binding to subsite ϩ2 difficult. Aromatic amino acids in the binding subsite ϩ2 can stabilize the transglycosylation receptor (33,34). In AaManA-ManIFG-M2, Trp-245 provided a hydrophobic sugar-binding platform in subsite ϩ2. However, AxMan113A lacked aromatic residues near this subsite. This special structural feature may result in the enzyme with no transglycosylation activity. Two different M2 electron-density maps were observed on the surface of AxMan113A-E223A-M2, which is far away from the catalytic sites (Fig. 4). A number of structural studies have revealed that carbohydrates are not only bound on carbohydrate-binding modules, but also observed on one or more surface-binding sites (SBSs). Compared with carbohydrate-binding modules, carbohydrates binding in SBSs are often through a  S8A). In chain B, Gln-58 and Asp-66 in a loop of AxMan113A, equivalent to Gly-66 and Asp-74 in AaManA, were flipped 90°, forming a "tweezer" binding linker that can target more mannose molecules efficiently (Fig. S8B) (37,38). It has been reported that a ␤-1,4-mannanase from Cryptopygus antarcticus (PDB entry 4OOZ) and several ␣-amylases have surface-binding sites (37)(38)(39). Moreover, mutations of the key amino acids near surface-binding sites decreased enzyme activity. To investigate the significance of these residues, site-directed mutagen- Structural analysis of GH family 113 ␤-1,4-mannanase esis was performed (Fig. S8C). The specific activity of six mutants (D66A, D73A, K76A, Q58A, N134A, and D65A) decreased, especially that of Q58A and D73A, which retained 56.8 and 82.7% of the initial enzyme activity, respectively. This indicated that these six residues play key roles in binding with ligands. The catalytic efficiencies (kcat/K m ) of Q58A and D73A toward M2 were decreased to 46.1 and 53.8% of WT, respectively. It could be deduced that residues around the surfacebinding sites have an effect on M2 affinity.
Rational design is an important method to modulate enzyme properties, in an attempt to understand structural information. Many GHs, such as ␤-gal, ␤-glycosidase, and ␣-L-arabinofuranosidase, have been modified through protein engineering (40 -42). However, protein engineering of GH family 113 ␤-1,4mannanases has never been reported. There were some differences in the structures of AxMan113A compared with other GH family 113 members. On the basis of the structural comparison of AxMan113A-E223A-M2 to AaManA-ManIFG-M2, three residues (Trp-239, Asn-237, and Lys-238) were identified in subsites ϩ1 and ϩ2 of AxMan113A, which had a unique amino acid arrangement (Fig. 6). Trp-239 in subsite ϩ1 had a larger side chain and extended into the groove, forming a semi-enclosed space, which impeded the accommodation of mannose units. However, Trp-239 formed hydrophobic interactions with Ϫ2 and ϩ2 mannoses. Thus, mutant W239Y was designed to reduce steric hindrance and retain the hydrophobic sugar-binding platform. W239A was designed to open the catalytic cleft. N237W not only retained the hydrogen bond, but also created an indole ring. K238A eliminated the side chain of Lys and partly opened the groove of subsite ϩ2. Mutant W239A lost most of its activity, confirming that the hydrophobic interaction is crucial for binding with ligands. However, ␤-1,4-mannanase activity of W239Y increased significantly to 136.4% of the WT. Mutant N237W enhanced affinity to the ϩ2 mannose, and enzyme activity was improved to 163.4%. The enzyme activity of K238A was improved to 212.9% of the WT (Fig. 6E). It has been reported that increasing intramolecular hydropho-bic interactions can improve structure stability (43,44). Herein, hydrophilic residue Asn-237 was mutated to hydrophobic tryptophan, forming a more favorable hydrophobic region with Trp-273, which is beneficial for substrate binding. Indeed, compared with WT, mutant N237W displayed the highest activity in acidic and higher temperature conditions (Fig. 7). According to the result of sequence alignment (Fig. 1), several polar residues (Cys, Ser, and Tyr) exist in six GH family 113 ␤-1,4-mannanases, but were replaced by nonpolar residue Val-139 in AxMan113A. Mutant V139C showed an acidic shift of 1.0 unit of optimal pH. In general, polar group burial makes a large contribution to structure stability (45,46). Mutating the nonpolar residue Val-139 to a polar residue Cys might be one reason for the optimal pH change. These results provided novel insights into the mechanism of GH family 113 members, which might have great significance in improving the hydrolysis efficiency of ␤-1,4-mannanases.
In conclusion, we performed structural and biochemical analyses to provide an in-depth description of the catalytic mechanism of GH family 113 members. Analysis of the complex structures revealed at least six binding sites (Ϫ4 to ϩ2) in the groove, but Lys-238 and Trp-239 were located near subsites ϩ1 and ϩ2, blocking the catalytic cleft. This prominent structural feature can mostly accommodate mannose units at subsites Ϫ4 to Ϫ1. Site-directed mutagenesis yielded a series of mutations designed to affect the enzyme's properties. Mutants K238A and W239Y successfully opened the catalytic cleft and their enzyme activities were improved to 212.9 and 163.4% of the WT, respectively. In addition, two M2-binding sites were discovered on the surface of AxMan113A, never before reported in GH family 113 members. Kinetic analysis of key residues near the surface sites suggested these sites' important role in substrate binding. The structural information of AxMan113A not only provides further insights into its catalytic mechanism, but also a direction for protein engineering in GH family 113 members.
The recombinant plasmid pET28a-AxMan113A was transformed into E. coli Rosetta (DE3) for expression. Seed cultures of E. coli harboring AxMan113A were incubated in Luria Bertani (LB) medium containing 50 g/ml of kanamycin at 37°C on a rotary shaker at 200 rpm to an A 600 of 0.6 -0.8. Isopropyl ␤-D-1-thiogalactopyranoside (1 mM) was added to induce expression at 30°C overnight. Cells were harvested by centrifugation at 10,000 ϫ g for 5 min. The precipitate was suspended in buffer A (20 mM imidazole, 20 mM Tris-HCl, pH 8.0, 500 mM NaCl) and disrupted by ultrasonication. The cell debris was centrifuged at 10,000 ϫ g for 10 min and the supernatant was loaded into a Ni-IDA column (1 ϫ 5 cm; GE Life Sciences) pre-equilibrated with buffer A. The column was washed with buffer A followed by buffer B (50 mM imidazole, 20 mM Tris-HCl, pH 8.0, 500 mM NaCl) and then eluted with buffer C (200 mM imidazole, 20 mM Tris-HCl, pH 8.0, 500 mM NaCl) at a flow rate of 1.0 ml/min. The eluted enzyme was subjected to a Sephacryl S-100 gel filtration column (1 ϫ 100 cm; GE Life Sciences), eluted with buffer D (20 mM Tris-HCl, pH 8.0, 100 mM NaCl) and concentrated to 10 mg/ml by ultrafiltration for crystallization. The concentration of the protein was measured by the Lowry method (47). All mutants were expressed and purified by the same method.

Enzyme characterization
Enzyme assay for AxMan113A was performed using the 3,5dinitrosalicylic acid (DNS) method (48) with 0.5% (w/v) LBG. Briefly, 100 l of suitably diluted enzyme was added to 900 l of LBG solution in 50 mM McIlvaine buffer (pH 6.5). After incubation at 45°C for 10 min, the reaction was terminated by adding 1 ml of DNS and boiling for 15 min. Enzyme activity in the reaction mixture was immediately measured at 540 nm after the addition of 1.0 ml of sodium potassium tartrate (40%, w/v). One unit of enzyme activity was defined as the amount of enzyme producing 1 mol of reducing sugars (mannose equiv-alents) per min. The optimal pH of AxMan113A was determined in various buffers (50 mM): McIlvaine buffer (pH 3.0 -7.0), citrate buffer (pH 3.0 -6.0), phosphate buffer (pH 6.0 -8.0), CHES (pH 8.0 -10.0), Gly-NaOH (pH 9.0 -10.5), and CAPS (pH 10 -11). The pH stability of AxMan113A was examined by determining the residual activity after incubation in these buffers at 30°C for 30 min. The optimal temperature was determined in the temperature range of 30 -70°C in 50 mM McIlvaine buffer (pH 6.5). The thermostability of AxMan113A was investigated by measuring residual activity after incubation at different temperatures for 30 min in the same buffer. All measurements were performed in triplicate.

Substrate specificity and hydrolysis properties
Substrate specificity toward 0.5% (w/v) of various mannans: konjac powder, LBG, and guar gum, was determined using the DNS method (48). HPLC with a refractive index detector was used to evaluate the substrate specificity of AxMan113A toward 0.1% (w/v) manno-oligosaccharides M2, M3, M4, M5, and M6 (49). Suitably diluted AxMan113A and manno-oligosaccharides were incubated at 45°C in 50 mM McIlvaine buffer (pH 6.5) for 10 min. A 10-l aliquot of the mixture was then injected into a Shodex sugar KS-802 (4.5 ϫ 250 mm) column and eluted by mobile phase (water) at a flow rate of 0.8 ml/min. The temperatures of column and detector were 35 and 60°C, respectively. One unit of enzyme activity was defined as the amount of enzyme that released 1 mol of manno-oligosaccharides/min. For determination of ␤-mannosidase activity of AxMan113A, 50 l of diluted enzyme was added to 200 l of 3.75 mM pNPM in 50 mM McIlvaine buffer (pH 6.5), then incubated at 45°C for 10 min. A total of 750 l of saturated sodium tetraborate was added to terminate the reaction. Enzyme activity in the reaction mixture was measured at 405 nm. One unit of enzyme activity was defined as the amount of enzyme liberating 1 mol of pNP/ min under the described conditions. The degradation of different mannans or manno-oligosaccharides by AxMan113A was detected by TLC. AxMan113A (5 units) was mixed into 1% (w/v) mannans or manno-oligosaccharides in 50 mM McIlvaine buffer (pH 6.5) at 45°C for 12 or 4 h. The hydrolysis of M2r, M3r, M4r, and M5r by AxMan113A were also analyzed. Purified AxMan113A (5 units) was incubated with 1% (w/v) M2r, M3r, M4r, and M5r under the same conditions for 4 h. Samples were withdrawn at the different time points and boiled for 10 min. The supernatants were spotted on a silica gel plate (Merck Silica Gel 60 F 254 , Germany) and developed in n-butyl alcohol: acetic acid:water (2:1:1, v/v). The plates were heated at 130°C for 5 min after spraying with methanol containing 2% (v/v) H 2 SO 4 .

Kinetic parameters for manno-oligosaccharides
The Michaelis-Menten parameters of AxMan113A for various manno-oligosaccharides (M2, M3, M4, M5, and M6) were determined by HPLC. Different concentrations of mannooligosaccharides were incubated with the appropriate amount of AxMan113A at 45°C in 50 mM McIlvaine buffer (pH 6.5) for 5 min. The mixtures were withdrawn followed by boiling for 5 min. The injection volume was 10 l and eluted by water at a flow rate of 0.8 ml/min. The temperatures of column and detec-Structural analysis of GH family 113 ␤-1,4-mannanase tor were 35 and 60°C, respectively. The catalytic efficiencies of mutants (Q58A and D73A) toward M2 were determined using the same method. The K m and V max values were calculated by nonlinear regression using Grafit software.

Crystallization and data collection
Crystallization experiments were performed in 48-well plates by sitting drop vapor diffusion at 293 K. The crystals were obtained from a drop containing 1 l of protein solution (10 mg/ml) and 1 l of reservoir solution (100 mM HEPES sodium, pH 6.8, 15% (v/v) 2-propanol, and 20% (w/v) PEG 4000). Initial crystals of AxMan113A were grown in 100 mM HEPES sodium (pH 7.5), 10% 2-propanol, and 20% PEG 4000. To obtain highquality crystals, we tried to optimize the initial screening conditions. Better crystals were obtained in the drops by mixing 2 l of protein solution and 1 l of reservoir solution at 293 K for 7 days. To obtain complex structures, 5% (w/v) M2, M3, M5, or M6 were added to the mutant AxMan113A-E223A. The mixtures were first incubated at 293 K for 2 h, and then screened under the same conditions as for the apo form.
For the X-ray diffraction experiments, a single crystal was fished out and immediately soaked in cryoprotectant solution (20% (v/v) glycerol under crystallization conditions), and then flash-cooled in liquid nitrogen. The X-ray diffraction data of AxMan113A were collected by beamline BL18U at the Shanghai Synchrotron Research Facility (SSRF, China). X-ray data for other complex crystals were obtained by beamline BL17U at SSRF. All data were indexed, integrated and scaled by the HKL-2000 package (50).

Structure determination and refinement
The structure of AxMan113A was determined by a molecular replacement method using AaManA (sequence identity 43%) as the search model. Phenix.autobuild was used for automatic building. The structures were completed with alternating rounds of manual model building with Coot (51) and Phenix.refine programs (52). The final model was analyzed by MolProbity (53). Structure illustrations were prepared with PyMOL. Ligplus (54) was used to analyze the protein-ligand interactions. The sequence alignments were created with ClustalX2 (55) and ESPript (56).
Author contributions-X. Y. and Y. L. investigation; X. Y. writingoriginal draft; Z. Q. and S. Y. data curation; Q. Y. and Z. J. supervision.