Identification of a Novel Di-D-Fructofuranose 1,2’:2,3’ Dianhydride (DFA III) Hydrolysis Enzyme from Arthrobacter aurescens SK8.001

Previously, a di-D-fructofuranose 1,2’:2,3’ dianhydride (DFA III)-producing strain, Arthrobacter aurescens SK8.001, was isolated from soil, and the gene cloning and characterization of the DFA III-forming enzyme was studied. In this study, a DFA III hydrolysis enzyme (DFA IIIase)-encoding gene was obtained from the same strain, and the DFA IIIase gene was cloned and expressed in Escherichia coli. The SDS-PAGE and gel filtration results indicated that the purified enzyme was a homotrimer holoenzyme of 145 kDa composed of subunits of 49 kDa. The enzyme displayed the highest catalytic activity for DFA III at pH 5.5 and 55°C, with specific activity of 232 U mg-1. K m and V max for DFA III were 30.7 ± 4.3 mM and 1.2 ± 0.1 mM min-1, respectively. Interestingly, DFA III-forming enzymes and DFA IIIases are highly homologous in amino acid sequence. The molecular modeling and docking of DFA IIIase were first studied, using DFA III-forming enzyme from Bacillus sp. snu-7 as a template. It was suggested that A. aurescens DFA IIIase shared a similar three-dimensional structure with the reported DFA III-forming enzyme from Bacillus sp. snu-7. Furthermore, their catalytic sites may occupy the same position on the proteins. Based on molecular docking analysis and site-directed mutagenesis, it was shown that D207 and E218 were two potential critical residues for the catalysis of A. aurescens DFA IIIase.


Introduction
In general, to grow in nature or in laboratory, the heterotrophic bacteria require carbohydrates as an energy source for cell growth. They secrete extracellular polysaccharide hydrolase to degrade polysaccharides into low-molecular-weight carbohydrates, especially monosaccharides, and further conduct the intracellular catabolism of the carbohydrate hydrolysates to obtain energy for growth. In nature, many polysaccharides can be used as an energy source, such as xylan [1], starch [2], mannan [3], and inulin [4]. and proposed that DFA IIIase hydrolyzes DFA III to inulobiose, inulobiose is further hydrolyzed into two fructose molecules by β-fructofuranosidase, and then fructose becomes the energy source for cell growth [26][27][28]. Thus far, gene cloning of DFA IIIase and characterization of the recombinant DFA IIIase have been only investigated by Saito et al. from Arthrobacter sp. H65-7 [29]. The DFA IIIase from Arthrobacter sp. H65-7 shows a relatively close relationship with IFTase (DFA I-forming) and IFTase (DFA III-forming) based on phylogenetic tree analysis [29]; in addition, it shows 44-47% of amino acid identity with all the reported IFTases [8].
In our previous studies, an IFTase (DFA III-forming)-producing microorganism, Arthrobacter aurescens SK8.001, was isolated using inulin as a sole carbon and energy source [30]. Specially, IFTase (DFA III-forming) was significantly induced when inulin was used as sole carbon source; Moreover, the gene encoding IFTase (DFA III-forming) from A. aurescens SK8.001 was sequenced (GenBank accession No. HM138085) and the extracellular overexpression was further studied in Escherichia coli [31]. When inulin was used as a sole carbon source for A. aurescens SK8.001 growth, high amounts of DFA III but no fructose were detected in the extracellular culture broth [23]; therefore, it was suggested that DFA III was first assimilated into the cells of A. aurescens and was probably converted by DFA IIIase to inulobiose for further catabolism. Regarding this probable metabolic pathway of DFA III, it is proposed that A. aurescens SK8.001 might have a DFA IIIase-encoding gene similar to that of Arthrobacter sp. H65-7. In the current article, the DFA IIIase-encoding gene from A. aurescens SK8.001 was obtained and sequenced, and the recombinant enzyme was identified and characterized, in addition, the potential residues of active site of DFA IIIase were first proposed based on homology modeling and molecular docking accompanied with site-directed mutagenesis. The DFA IIIase from A. aurescens SK 8.001 was abbreviated as AaDFA IIIase in this study.

Chemicals and reagents
The resin for protein purification, the Chelating Sepharose Fast Flow, was obtained from GE (Uppsala, Sweden). The standard of DFA III was purchased from Wako Pure Chemical Industries (Osaka, Japan). Electrophoresis reagents were purchased from Bio-Rad (Hercules, CA, USA). Isopropyl-β-D-1-thiogalactopyranoside (IPTG) and all chemicals used for enzyme assays and characterization were at least of analytical grade obtained from Sigma (St. Louis, MO, USA) or Sinopharm Chemical Reagent (Shanghai, China).

Plasmids, bacterial strains, and culture conditions
The pET-22b(+) expression vector was obtained from Novagen (Darmstadt, Germany). A. aurescens SK 8.001 was isolated by our laboratory previously from soil and kept in China Center for Type Culture Collection (CCTCC) under the accession number of M 207185. The strain was grown at 30°C and 200 rpm for 12 h in flask containing culture medium (peptone 5 g L -1 , NaCl 5 g L -1 , and yeast extract 3 g L -1 , pH 7.0). E. coli strains DH5α and BL21(DE3) were obtained from Sangon Biological Engineering Technology and Services (Shanghai, China), which were used as the host strains for DNA cloning and expression, respectively. The recombinant E. coli strains were routinely grown in Luria-Bertani (LB) medium with ampicillin (100 μg mL -1 ) in a rotary shaker at 37°C and 200 rpm.
The complete genome sequence of A. aurescens TC1 has been released in GenBank (NCBI accession number: CP000474.1), and revealed that a putative IFTase gene was included in the genome (gene locus_tag: AAur_0696; protein ID: ABM07149.1). The sequence of the oligonucleotide primers used for gene cloning was based on the putative gene. Forward (5'-CGCCAT ATGGAAATTGACGAGACGCTT-3') and reverse (5'-TATACTCGAGTCGAGGAATACC CGCGCC) primers were designed to introduce the NdeI and XhoI restriction sites (underlined). The polymerase chain reaction (PCR) was performed by Taq Plus DNA polymerase for 35 cycles consisting of 94°C for 30 s, 57°C for 1 min, and 72°C for 1 min, followed by a final extension step at 72°C for 10 min. The amplified DNA fragment (1.4 kb) was extracted using the SanPrep column DNA gel extraction kit and was sequenced by Sangon Biological Engineering Technology and Service (Shanghai, China). The target fragment was fused with a 6× histidine-tag at the C-terminal, digested using NdeI and XhoI, and then was inserted into the same restriction sites of pET-22b(+) plasmid.
The reconstructed plasmid was transformed into E. coli BL21(DE3) for overexpression of the recombinant DFA IIIase. For protein expression, the recombinant E. coli BL21(DE3) was cultivated in 500 mL LB medium supplemented with 100 μg mL -1 ampicillin in a rotary shaker at 37°C and 200 rpm. When the absorbance (600 nm) of the bacterial culture reached 0.6, IPTG was added at a final concentration of 0.5 mM and the cells were grown at 28°C for an additional 6 h.

Purification of the recombinant DFA IIIase
The cells were collected by centrifugation at 10,000 × g (4°C, 10 min), washed with phosphate buffer (pH 7.0, 50 mM), and disrupted by sonication at 4°C for 6 min (pulsations of 3 s, amplify 90) using a Vibra-Cell TM 72405 Sonicator (Sonics, Newtown, CT, USA). The lysates were centrifuged at 20,000 × g (4°C, 30 min) to remove the cell debris. As fused with a 6×histidine tag, the expressed enzyme was allowed to be purified by nickel-affinity chromatography. The supernatant was loaded onto a Ni 2+ -chelating Sepharose Fast Flow column (GE Healthcare, Uppsala, Sweden) equilibrated with a binding buffer (50 mM sodium phosphate buffer, 500 mM NaCl, pH 7.5). The unbound proteins were eluted from the column using a washing buffer (50 mM sodium phosphate buffer, 500 mM NaCl, 50 mM imidazole, pH 7.5), and then the recombinant enzyme was eluted by using a elution buffer (50 mM sodium phosphate buffer, 500 mM NaCl, 500 mM imidazole, pH 7.5). To remove imidazole, the collected fractions were dialyzed against sample buffer (50 mM citrate buffer, pH 5.5). After dialysis, the resulting solution was used as the purified enzyme for further studies.

Protein concentration, electrophoresis, and gel filtration chromatography
According to the method of Bradford [32], the protein concentration was determined using bovine serum albumin as a standard. The molecular mass and purity of the purified A. aurescens DFA IIIase were examined by denaturing discontinuous sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) on a 5% stacking gel and a 12% separating gel as described by Laemmli [33]. Gels were stained with Coomassie Brilliant Blue R250 and destained with an aqueous mixture of 10% (V/V) methanol/10% (V/V) acetic acid at room temperature. Native molecular weight was estimated by gel filtration using high-performance liquid chromatography (HPLC) at a flow rate of 1 mL min -1 (column: TSK G2000SWxl, Tosoh Bioscience LLC, Tokyo, Japan; mobile phase: 0.1 M phosphate buffer, pH 6.7, containing 0.05% (W/V) NaN 3 and 0.1 M Na 2 SO 4 ; detection: UV at 280 nm).

Enzyme assay
The enzyme activity was measured by the determination of the amount of produced inulobiose from DFA III. The enzyme reactions were carried out at 55°C for 10 min in a 1 mL reaction mixture containing DFA III (10 g L -1 ), citrate buffer (50 mM, pH 5.5), and 100 nM enzyme, and stopped by boiling for 10 min. The mixture was centrifuged for 30 min at 20,000 × g, and the supernatant was filtered through a 0.22 μm membrane filter into a sample vial for HPLC analysis. The carbohydrates were determined by HPLC (Agilent 1200 system, Agilent technologies, Santa Clara, CA, USA) equipped with a refractive index detector and an Asahipak NH2P-5004E column (4.6 mm × 250 mm, Shodex, Tokyo, Japan; column temperature: 25°C; mobile phase: 65% (V/V) acetonitrile; flow rate: 1 mL min -1 ). One unit of DFAIIIase was defined as the amount of enzyme that produces 1 μmol inulobiose per min at pH 5.5 and 55°C.
Determination of the reaction product from DFA III by the purified DFA IIIase For preparation of reaction products, the purified DFA IIIase solution (100 nM) and 100 g L -1 (W/V) DFA III were mixed in a 100 mL enzyme reactor and maintained at 55°C and pH 5.5 for 2 h. The reaction temperature was controlled with circulating water bath and the mixture was heated at 100°C for 10 min to terminate the enzyme reaction. Thereafter, the mixture was centrifuged at 20,000 × g (4°C, 30 min) and supernatant was filtrated through a 0.22 μm membrane filter. The filtrate was then applied to a preparative column (10.0 × 300 mm, Sepax, Newark, Delaware, USA) with HPLC (Agilent 1200 system) equipped with a refractive index detector (temperature: 80°C, mobile phase: 2.5 mM trifluoroacetic acid; flow rate: 0.5 mL min -1 ; injection volume: 200 μL). The fraction of product was collected and pooled in a 50 mL centrifuge tube and lyophilized with a vacuum freezing dryer. The dried product was dissolved in deuteroxide (D 2 O) and analyzed by 13 C-NMR system 300 spectrometer (Varian, Palo Alto, CA, USA) using 3-trimethylsilyl-1-pro-panesulfonic acid sodium salt (DSS) as a standard.
In addition, the ultra-performance liquid chromatography and electrospray ionization quadrupole time-of-flight mass spectrometry (UPLC/ESI-QTOF-MS, SYNAPT™ High Definition Mass Spectrometry™ system; Waters, Milford, MA, USA) was used for the product analysis. The system equipped with a TSK-gel Amide-80 column (4.6 mm id × 25 cm, 5 μm, 80 Å, Tosoh, Tokyo, Japan) was operated in the negative ion mode with a negative electrospray ionization source (ESI -), and the cone voltage, capillary voltage, source temperature, and mass range were set as 30 V, 3.0 kV, 100°C, and 50-1000 m/z, respectively.

Effect of pH and temperature on the enzyme activity
The optimum pH was determined using two buffer systems with different pH values at 55°C for 10 min, including sodium citrate buffer (50 mM, pH 4.0-5.5) and sodium phosphate buffer (50 mM, pH 6.0-8.0). The effect of temperature on the activity of the AaDFA IIIase was analyzed by assaying the enzyme samples over the range of 30-60°C at pH 5.5 for 10 min.

Determination of kinetic parameters
To determine kinetic parameters of the AaDFA IIIase, various concentrations of DFA III (2.0-50 mM) in 50 mM of citrate buffer (pH 5.5) was incubated with the enzyme at 55°C for 10 min, and the reactions were stopped by heating in boiling water for 10 min. The inulobiose produced was measured by HPLC assay method. The Michaelis-Menten constant (K m ) and the maximum rate (V max ) values were estimated by a nonlinear regression method.

Molecular modeling and docking
To obtain a template of molecular modeling, the amino sequence of AaDFA IIIase was submitted to SWISS-MODEL protein-modeling server (http://www.expasy.ch/swissmod/ SWISS-MODEL.html) with Automated Mode [34][35][36][37], and the crystallographic structure of BsIFTase from Bacillus sp. snu-7 (PDB ID: 2INU) was selected as the template of homology modeling for AaDFA IIIase and its mutants. Stereochemical quality of the modeled structures were assessed by analyzing Ramachandran map using WinCoot, the compatibility of an atomic model (3D) with its own amino acid sequence (1D) was determined by VERIFY-3D (a structure evaluation server) [38] through ASVES server (http://services.mbi.ucla.edu/SAVES/). Molecular docking of DFA III to AaDFA IIIase and mutants models was carried out with the Autodock4.2 software package [39]. The structure of ligand DFA III was obtained from Pub-Chem (http://pubchem.ncbi.nlm.nih.gov/) (CID of DFA III: 196181), and this structure was transformed to pdb format with the GlycoBioChem PRODRG2 Server [40]. Proteins were added with polar hydrogen atoms using the Hydrogen module in AutoDock Tools (ADT), and all the torsion angles in the DFA III were set free to perform flexible docking. The empirical free energy function and Lamarckian genetic algorithm (LGA) were used for docking with the following settings: a population size of 150 dockings and 5 million energy evaluations, a maximum number of 27,000 generations, a mutation rate of 0.02, a crossover rate of 0.80 and an elitism value (number of top individuals that automatically survive) of 1. Fifty independent docking runs were carried out for each ligand. Results were clustered according to the root-mean-square deviation (RMSD) criterion. The best docked conformations were selected as the initial conformations to evaluate the molecular docking between the enzyme and the ligands. The figures were prepared using Discovery Studio Visuallizer 4.0 (Accelrys, San Diego, USA)/Pymol.

Mutagenesis
Site-directed mutagenesis of the AaDFA IIIase gene of A. aurescens SK8.001 was implemented by one-step PCR method, the aforementioned recombinant plasmid harboring the wild-type AaDFA IIIase gene was used as template and a pair of synthetic complementary oligonucleotides were used as primers. Oligonucleotides shown in Table 1 were synthetized by Shanghai Generay Biotech Co., Ltd (Shanghai, China) and the underlined sequences indicated mutated codons. The amplified PCR products were digested with DpnI restriction enzyme and then transformed into E. coli DH5α host cells. The nucleotide sequences of mutants were verified by DNA sequencing. For expression of mutant enzymes, the plasmids containing the various mutant gene of AaDFA IIIase were introduced into E.coli BL21(DE3) host cells. The purification and enzyme assay of these mutants were performed as those of the wild-type AaDFA IIIase.  Fig 2 shows a phylogenetic tree of DFA IIIases and all reported IFTases. The dendrogram indicates that DFA IIIases and IFTases from Arthrobacter genus form two phylogenetically independent groups and that IFTases are further split into two independent groups (DFA I-forming and DFA III-forming IFTase groups). All these divergences probably indicate that IFTases and

Purification of the AaDFA IIIase
The analysis of the deduced amino acid sequence using SignalP 4.1 Server (http://www.cbs.dtu. dk/services/SignalP/) revealed that there was no signal peptide in AaDFA IIIase. Due to fusing with a 6×histidine-tag (as mentioned in Materials and Methods), the recombinant AaDFA IIIase was allowed to be purified from cell-free extract through a Ni 2+ -chelating affinity chromatography column by one step. Shown in Fig 3A, the enzyme was purified to electrophoretic homogeneity with an estimated molecular mass of 49 kDa, which agreed well with 49,879 Da calculated based on the deduced amino acid sequence. The gel filtration showed that the native purified enzyme had a total molecular mass of approximately 145 kDa (Fig 3B). These results suggested that the AaDFA IIIase is a trimer with three identical subunits. By contrast, the DFA IIIase from Arthrobacter sp. H65-7 is a dimer, and its molecular mass, as assayed by SDS-PAGE, was 61 kDa while native total molecular weight as determined by gel filtration was 125 kDa. Previously, the crystal structure of the IFTase (DFA III-forming) from Bacillus sp. snu-7 (BsIFTase) revealed that trimerization is a prerequisite for the catalytic reaction as the active site is located at the monomer-monomer interface. DFA IIIases (EC 3.2.1.-) and IFTases (EC 4.2.2.17, and EC 4.2.2.18) were categorized as glycoside hydrolase (GH) family 91 and had a relatively high sequence identity [42]; therefore, it was suggested that the oligomerization is probably essential for the catalytic reaction of AaDFA IIIase as BsIFTase.

Identification of Reaction Products
To determine the reaction product, a preparative column was used to separate the product from reaction mixture. There was only one product in the reaction mixture (data not shown), which was analyzed by both NMR and LC-ESI-MS. As shown in the S2 Fig, the purified product exhibited the (M-H)ion peak of m/z 341 Da, indicating that the molecular weight should be 342 Da. In addition, the 13 C-NMR spectrum (shown in the Table 3) indicated that the chemical shift agreed well with those of reducing and non-reducing terminal fructose of inulobiose. Hence, the product was identified as the inulobiose. Moreover, given the chemical shift of C2 of the reducing terminal fructose, the inulobiose was determined as 1-O-β-D-fructofuranosyl-D-fructopyranose according to the method described by Sakurai et al. [26]. Therefore, the  inulobiose produced by AaDFA IIIase in this work had the same chemical structure as Arthrobacter sp. H65-7 DFA IIIase, but was different from A. ureafaciens DFA IIIase, which was 1-Oβ-D-fructofuranosyl-D-fructofuranose [27]. These results indicated that the recombinant enzyme from A. aurescens SK8.001 characterized herein should be a DFA III hydrolysis enzyme, DFA IIIase. Then, why does the A. aurescens SK8.001 contain a DFA IIIase-encoding gene? It is suggested that the DFA IIIase is probably involved in the catabolic pathway of inulin in A. aurescens SK8.001. In our previous studies, A. aurescens SK8.001 grew very well, accompanied by the biological accumulation of DFA III in broth, when using inulin as the sole carbon source [23]. After DFA III production by the IFTase (DFA III-forming), the strain might need to obtain energy through DFA III hydrolysis by DFA IIIase and further catabolic reactions [8]. So far, there have been very few studies focusing on microbial inulin catabolism via DFA. In 1975, Tanaka et al. reported that the enzymes from autolysate of A. ureafaciens converted DFA III to fructose with inulobiose as an intermediate [27]. The DFA IIIases from A. ureafaciens and Arthrobacter sp H65-7 were identified to be able to hydrolyze DFA III by DFA IIIase to inulobiose [26,28]. In addition, Matsuyama et al. found that DFA I and inulobiose could be reversibly converted [43]. For inulin-assimilation though DFA III as intermediate, Saito et al. proposed that inulin is firstly depolymerized to the DFA III by IFTase (DFA III-forming), DFA III is further hydrolyzed to inulobiose by DFA IIIase, and then inulobiose is hydrolyzed to two fructose molecules to enter the tricarboxylic acid (TCA) cycle for generating energy [29]. The enzyme hydrolyzing inulobiose into fructose has not been identified yet, but was proposed to be a type of β-fructofuranosidase [29].
Effect of pH and temperature on the AaDFA IIIase activity Shown in Fig 4A, the enzyme exhibited a relatively wide pH spectrum (from 5.0-7.0, approximately above 80% relative activity), with the highest activity at pH 5.5. The optimum pH (pH 6.0) was close to that of DFA IIIase from Arthrobacter sp. H65-7 [26]. The enzyme showed the highest activity at 55°C and the activation energy for the hydrolysis reaction calculated from Arrhenius equation was 47.8 kJ mol -1 (Fig 4B). The maximum enzymatic activity of 232 ± 15 U/mg was achieved at pH 5.5 and 55°C. By comparison, optimum pH and temperature of the DFA IIIases from A. ureafaciens ATCC21124 were pH 6.5 and 60°C [28], while the enzyme from Arthrobacter sp. H65-7 showed the highest activity at a lower temperature (45°C) [26].

Enzyme kinetics
The K m and V max were determined by using nonlinear regression plots in three independent experiments, and the values were 30.7 ± 4.3 mM and 1.2 ± 0.1 mM min -1 , respectively (S3 Fig). The K m of AaDFA IIIase for DFA III was similar to the one from A. ureafaciens ATCC21124 (31 mM) [28] but higher than the one from Arthrobacter sp. H65-7 (12.5 mM) [26].

Homology building
So far, the structure of DFA IIIase has not been determined yet. The amino acid sequence of AaDFA IIIase was submitted to SWISS-MODEL with the Automated Mode and compared to proteins with known structure information. The results revealed that AaDFA IIIase had the highest identity (50.88%) with IFTase (DFA III-forming) from Bacillus sp. snu-7 (PDB ID: 2INU) but shared less than 30% identity with the others. A model of tertiary structure of AaDFA IIIase was constructed based on the crystallographic structure of Bacillus sp. IFTase (as shown in Fig 5A). The generated model was then validated by Ramachandran plot. As shown in S4 Fig, 90.79% of amino acid residues were located in preferred regions, 4.72% of the residues were located in allowed regions, and 4.49% of the residues were located in outlier regions of the Ramachandran plot. VERIFY-3D analysis showed that 83% of the residues had an averaged 3D-1D score 0.2. The models of various mutants had similar results. These results indicated that the predicted models were acceptable.
The overall structures of AaDFA IIIase and its mutants were right-handed parallel β-helix structures, which were the same as that of the IFTase reported by Jung [42]. Fig 5B showed that the monomer of AaDFA IIIase model superimposed with the majority of regions of IFTase monomer except two long loop regions. It indicated that the trimerization was probably also a prerequisite for catalytic activity of AaDFA IIIase and that the active site might be located at the monomer-monomer interface as IFTase. Based on this speculation, the constructed model of AaDFA IIIase was superimposed onto the template crystal structure (shown in Fig 6A), and the result showed that, except H287, all the residues of AaDFA IIIase corresponding to the active positions in BsIFTase were completely the same as those of BsIFTase, indicating that these residues in AaDFA IIIase probably form an active pocket that recognizes and catalyzes DFA III.

Molecular docking and site-directed mutagenesis
Molecular docking of DFA III into the putative active pocket of AaDFA IIIase was performed by AutoDock4.2. The best ligand docking conformation was selected based on the docking energy and root mean square deviation (RMSD) criterion. As shown in Fig 6B, several hydrogen bonds were generated between DFA III and residues of the side-chains of the AaDFA IIIase model. Jung et al. reported that E244 in BsIFTase acts as a general base to activate the acceptor and that D233 orients the donor molecule into an optional configuration for catalysis, suggesting that D233 and E244 play critical roles in catalyzing the biological reaction [42]. Shown in  Fig 6B, the two residues D207 and E218, corresponding to the D233 and E244 in BsIFTase, also directly interacts with the substrate DFA III. To determine the functional roles of these two residues, site-mutagenesis was performed, and the mutant enzymes were expressed and purified as the wild-type AaDFA IIIase described above. Each of the purified proteins appeared as a single major band on SDS-PAGE (Fig 6C). The enzymatic activities of various mutants, including D207A, D207N, E218A, and E218N, were measured; however, all of them were completely inactive. Therefore, it was suggested that the residues at position of 207 and 218 were indispensable for the catalytic activity of AaDFA IIIase.
The molecular docking analyses of mutants and DFA III were implemented as the wild-type AaDFA IIIase. For D207 mutants (shown in Fig 7A and 7B), E218 still interacted with DFA III molecule, and no hydrogen bond was generated between D207A and DFA III, while the amino-group of D207N hydrogen bonded with DFA III. For E218 mutants (Fig 7C and 7D), the alanine residue in the mutant E218A also had no hydrogen bond with DFA III, whereas the amino group of E218N hydrogen bonded with DFA III as D207N. Moreover, D207 formed a weak hydrogen bond with DFA III at a distance of 3.2 Å (Fig 7C). The negatively charged residues such as aspartate and glutamate usually act as general bases or nucleophiles. The mutagenesis of these residues probably changes the charge state and inactivates the enzyme, which have been confirmed by many enzymes such as IFTase [42], levan fructotransferase [44] and other glycosyltransferases [45,46]. Furthermore, multiple sequence alignment of the DFA IIIases and the reported IFTases (DFA I-forming and DFA III-forming) were carried out, as shown in S5 Fig. It is interesting that all residues responding to D207 and E218 of AaDFA IIIase are completely conserved in all sequences. Eleven non-catalytic critical residues of BsIFTase, including S133, I134, R174, Y197, F290, P291, R292, Q256, Q313, F315, and F346, were also highly conserved in these enzymes. It suggests that all of the DFA IIIases and IFTases probably share very similar active pocket and catalytic residues. To evaluate this hypothesis, the comparison of these three-dimensional structures should be performed. However, only the crystallographic structure of BsIFTase has been reported. Hence, the three-dimensional structures of IFTase (DAF I-forming) and DFA IIIase should be further investigated.

Author Contributions
Conceived and designed the experiments: SY XW TZ TS LF BJ WM. Performed the experiments: SY XW. Analyzed the data: SY XW LF BJ WM. Contributed reagents/materials/analysis tools: SY XW BJ WM. Wrote the paper: SY WM.