Over-expression and purification of BCAI
The araA gene contained an open reading frame (ORF) of 1422 base pairs encoding a protein of 473 amino acids. Protein sequence alignments showed that, among the AIs with high activities toward D-galactose, BCAI was mostly similar to Lactobacillus sakei AI and Pediococcus pentosaceus AI with identities of 68.8 and 67.5 % respectively. By contrast, the identities to other AIs without obvious D-galactose activities from Bacillus strains including B. halodurans (57.1 %) and B. substilis (55.6 %) were lower (Fig. 2). Expression was induced upon addition of IPTG in Luria Bertani (LB) medium. The activity of BCAI crude extract was 2.3 U mg-1 for L-arabinose and 0.3 U mg-1 for D-galactose. Then, the crude extract was subjected to heat treatment (60 °C) and purified by HisTrap HP 5 mL column. SDS-PAGE showed distinct bands with a molecular mass around 55 kDa (expected size: 53.5 kDa, Additional file 1: Figure S1). The electrophoretically pure AI showed a specific activity of 8.0 U mg-1 toward L-arabinose (Table 1) and 1.1 U mg-1 toward D-galactose. BCAI was highly similar to the AI from B. coagulans 2-6 with an identity of 96.0 %, but no reports indicated that B. coagulans 2-6 AI possessed the same D-galactose activity .
Effects of temperature and pH on activity of BCAI
The effects of temperature were determined at 40 to 90 °C and pH 7.5 (Fig. 3a). BCAI displayed its maximal activity at 60 °C and retained above 85 % of the activity at 50 to 70 °C. Even at 80 to 90 °C, it still preserved above 60 % of the maximal activity. Compared with the AIs from B. stearothermophilus IAM11001, Lactobacillus fermentum CGMCC2921 and T. mathranii, BCAI was less sensitive to temperature change and could adapt to a broader range of temperatures [16, 17].
To investigate the effect of pH, enzyme assays were carried out at a series of pH from 2.2 to 9.0. The relative activity of BCAI reached the maximal value at pH 7.5 (Fig. 3b) and decreased by less than 10 % at pH 8.0 to 9.0. By contrast, the activity was weaker at acidic conditions. It decreased severely when the pH dropped to 5.0 as most of moderate alkaline AIs did previously, possibly because some side chain groups close to its substrate binding sites were difficult to ionize under this condition .
It has been experimentally proved that the optimum pH (pHopt) of AIs is affected by some crucial residues with polar groups, for example the E268 residue of B. halodurans AI (BHAI, pHopt =8.0) and the equivalent D269 of L. fermentum AI (LFAI, pHopt =6.5) [13, 17]. Modifications of the two residues to lysine (K) resulted that the pHopt of BHAI and LFAI decreased to 7.0 and 5.0, respectively [19, 20]. Protein sequence alignment showed D268 in BCAI was the counterpart of E268 of BHAI and D269 of LFAI (Fig. 2a). It could be presumed that if the D268 residue was changed to lysine, the pHopt of BCAI would probably decrease to a lower value.
Effects of metallic ions on activity and thermostability of BCAI
After BCAI was treated by ethylenediamine tetraacetic acid (EDTA), a dramatic loss (60 %) of activity was observed in an enzyme assay at 60 °C and pH 7.5. The EDTA-treated enzyme was then incubated in Tris-HCl (pH 7.5) solutions with different types of divalent metallic ions (Mg2+, Ca2+, Mn2+, Fe2+, Co2+, Ni2+, Cu2+). Enzyme assays showed that, except Cu2+, all other divalent ions could serve as activators (Fig. 4a). 0.5 mM Mn2+ and 0.5 mM Co2+ respectively boosted the activity by 270 and 190 % and the combination of them finally resulted in a 370 % increase. Previous researches had demonstrated that Mn2+ or Co2+ could assist AIs to transfer to correct substrate-binding conformations at elevated temperatures . They significantly boosted the activities of AIs from other strains such as T. maritima, L. fermentum CGMCC2921 and G. thermodenitrificans [9, 17, 22], but some of them required higher Mn2+ or Co2+ concentrations (≥2 mM). In this study, when the Mn2+ concentration was increased from 0.5 mM to 4 mM, the BCAI activity changed very slightly (Fig. 4b). 0.5 mM Mn2+ was completely sufficient for BCAI to maintain a high activity. Further increase of Mn2+ by many times could not boost BCAI activity effectively but also would influence the quality of products. For purified BCAI without EDTA treatment, which will be called raw purified BCAI below, the activity was 92.6 and 72.5 %, i.e. very close to the activity of EDTA-treated enzyme measured in 0.5 mM Co2+ and Mn2+ solutions (Fig. 4a), possibly because it originally bonded metal ions or other types of ligands in the E. coli cells.
As EDTA treatment would not be used in practical applications of AIs, the thermal stability test was carried out using the raw purified BCAI. Figure 5a showed it was perfectly stable at 60 °C. After an incubation of 2 h at this temperature, it preserved 90 % of the initial activity. When incubated at 70 °C or higher temperatures, the enzymatic activity declined very quickly by more than 80 % during the first 30 min. Some reports showed that adding Mn2+ could enhance the thermostability of EDTA-treated AIs [8, 9]. But in this study, a low concentration of Mn2+ did not make the same enhancement effect on the raw purified BCAI. As shown in Fig. 5b, after incubation with 0.5 mM Mn2+ at temperatures between 60 and 90 °C for 2 h, no positive changes appeared on the residual activities of the enzyme.
From the two perspectives of enzyme activity and thermostability, it seemed that external Mn2+ was not essential for the raw purified BCAI. Since a low amount of metallic ions can increase the purity and safety of products, BCAI will show its unique value in food-grade D-tagatose production.
Kinetic parameters of BCAI and molecular docking studies
Kinetic constants of BCAI were measured by following the rules of Lineweaver-Burck plots (Table 2). The catalytic efficiencies (k
m) of BCAI were 8.7 mM-1 min-1 and 1.0 mM-1 min-1 for L-arabinose and D-galactose respectively. According to the ratio (8.7: 1) between the k
cat/Km for L-arabinose and D-galactose, L-arabinose was obviously a more favorable substrate than D-galactose for BCAI. On the other hand, the k
cat/Km for D-galactose of BCAI was noticeable because many other AIs from Bacillus strains had not been reported to show any catalytic efficiency for D-galactose [5, 13, 14].
Molecular modelling and docking techniques were used for gaining a deeper understanding of the substrate specificity. In previous studies, the monomers of B. licheniformis AI (BLAI) and P. pentosaceus PC-5 AI (PPAI) had been comparatively modeled by using the crystal structure of E. coli AI as a template [23, 24]. However, the crystallographic analysis on L. fermentum CGMCC2921 AI (LFAI) and E. coli AI (ECAI) indicated that they were hexamers [25, 26]. Native-PAGE showed that the total molecular mass of BCAI was around 300 kDa, which meant that BCAI was also a hexamer (expected size: 320 kDa, Additional file 1: Figure S2). It could be assumed that substrate catalyses of these enzymes might be affected by subunit interactions. Therefore, monomer models were not accurate enough when protein subunit interactions were considered. In this study, half of the BCAI structure (a trimer, Additional file 2) was constructed on the base of LFAI (PDB ID: 4LQL, identity: 59.5 %) and ECAI (PDB ID: 4F2D, identity: 44.0 %) crystal structures. The present choice represents a compromise for achieving better conformations of the active sites and reducing computation size as well (Fig. 6a). The superimposition of the obtained BCAI structure with 4LQL and 4F2D ensured the conservation of the putative catalytic amino acids (E306, E331, H348 and H447) (Fig. 6b). The structure energy was sufficiently minimized through a 1000-step Conjugate Gradient Descent until the RMS Gradient reached 0.1.
Although molecular docking studies had been implemented on BLAI before, the result was not very representative because BLAI did not possess any D-galactose activity in experiments. Most AIs such as BCAI were able to catalyze L-arabinose and D-galactose simultaneously. As shown in Fig. 6c, two hydrogen bonds (2.15 Å, 2.34 Å) existed between the C2 hydroxyl group of L-arabinose and the oxygens of E306. One hydrogen bond (1.96 Å) was found between the C1 hydroxyl group and the oxygen of E331. According to the presumed catalytic mechanism of E. coli AI and B. licheniformis AI [26, 27], L-arabinose was firstly transformed to an enediol intermediate and then L-ribulose was formed. During the first step, protons were transferred through C1 and C2 of the substrate with the assistances of E306 and E331. The docking result here confirmed that E306 and E331 of BCAI played important roles of targeting the C2 and C1 hydroxyl groups of L-arabinose. The hydrogen bond interactions between L-arabinose and the residues were sufficiently strong. By contrast, the interactions for D-galactose were weaker. Among the two hydrogen bonds (2.35 Å, 1.94 Å) found in Fig. 6d, only the bond between the C3 hydroxyl group of D-galactose and E306 could promote the reaction according to the putative mechanism exposed above. E331 residue did not orientate the C1 hydroxyl group of D-galactose correctly, which would cause difficulties on proton transfer and slowed formation of enediol intermediate. This could be an explanation for why the formation of D-tagatose was always slower than that of L-ribulose in most AI catalyses.
Meanwhile, the C-DOCKER energies for the docking poses of L-arabinose and D-galactose were -9.39 kcal/mol and -7.07 kcal/mol respectively. A lower value indicates a more favorable binding, thus further confirming that D-galactose is poorer fit in the active site pocket of BCAI than L-arabinose.
Conversion of D-galactose to D-tagatose by using whole cells of recombinant E. coli
Since BCAI could isomerize D-galactose to D-tagatose, the feasibility of D-tagatose production was further studied. It was complicated to use purified enzyme as biocatalyst in industry. Instead, whole cells of recombinant E. coli was constructed and selected as a suitable biocatalyst for D-tagatose production.
To increase the efficiency of D-tagatose production, the biocatalytic conditions were optimized. The effect of cell concentration on D-tagatose production was firstly investigated. As shown in Fig. 7a, the highest conversion rate was obtained at 4.8 g DCW L-1. Then, the effect of reaction temperature was evaluated within a temperature range from 40 °C to 80 °C. Figure 7b showed that the conversion rate reached a maximum (40.8 %) at 60 °C. It was consistent with the optimum temperature of purified BCAI. Although E. coli cells suffer from viability loss and cellular structure transition when the temperature is higher than 55 °C , these serious damages to E. coli cells did not suppresses the transportation of D-galactose and D-tagatose at 60 °C. The effect of substrate concentration was tested in the range from 50 g L-1 to 250 g L-1 D-galactose (4.8 g L-1 cells and 60 °C). With the increase of D-galactose concentration, the amount of D-tagatose kept rising without obvious substrate inhibition, whereas the conversion rate decreased from 31.4 to 17.0 % (Fig. 7c). In order to achieve a higher yield of D-tagatose, 250 g L-1 galactose was firstly selected for the following experiment. Meanwhile, 150 g L-1 D-galactose was contained, because a relatively higher conversion rate was also expected.
Based on the above experiments, the time course of tagatose production at 150 g L-1 and 250 g L-1 galactose were performed under the optimal conditions. After 32 h biotransformation, the concentrations of D-tagatose were 48.1 g L-1 and 55.5 g L-1 respectively (Fig. 7d). The conversion rates were 32.1 and 22.2 % respectively. During the next 16 h, the conversion rate for 250 g L-1 D-galactose rose up a little to 27.2 %. The achieved conversion rates were attractive for industrial D-tagatose production. Although immobilized AIs from G. stearothermophilus, T. mathranii and T. neapolitana [28–30] had been used to produce D-tagatose before, the process in this study was easier to operate and due to the enzyme purification and immobilization steps were eliminated. It was a one-pot bioconversion process and introductions of high cell density cultivation and continuous reactors could hopefully improve its feasibility in the future.